Vrije Universiteit Brussel

A Method of Reaching Consensus with the Multi-Actor Multi-Criteria Analysis (MAMCA) methodology Huang, He; Macharis, Cathy; De Smet, Yves; Doan Nguyen, Anh Vu

DOI: 10.4028/www.scientific.net/KEM.807.165

Publication date: 2019

Document Version: Final published version

Link to publication

Citation for published version (APA): Huang, H., Macharis, C., De Smet, Y., & Doan Nguyen, A. V. (2019). A Method of Reaching Consensus with the Multi-Actor Multi-Criteria Analysis (MAMCA) methodology. 165-170. Abstract from 25th International Conference on Multiple Criteria Decision Making, Istanbul, Turkey. https://doi.org/10.4028/www.scientific.net/KEM.807.165

General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal Take down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Download date: 04. Oct. 2021

Table of Contents

Welcome Messages ...... 1

Message from Dr. Murat Köksalan President, International Society on MCDM ...... 1

Message from Dr. İlker Topcu Conference General Chair, MCDM 2019 ...... 3

General Information ...... 6

Venue ...... 6

How to reach campus ...... 6

Getting around in Istanbul ...... 7

Registration Desk ...... 7

Registration Fee ...... 8

Internet Access ...... 8

Lunch and Coffee Breaks ...... 8

Important Phone Numbers ...... 8

Disclaimer ...... 9

MCDM 2019 ...... 10

Committees ...... 10

Awards ...... 11

Plenary Sessions ...... 13

Tutorial Sessions ...... 14

Parallel Sessions ...... 14

Session Identifiers ...... 16

Instructions for Session Chairs ...... 16

Instructions for Presenters ...... 16

Publications ...... 17

Social Program ...... 19

MCDM 2019 in Numbers ...... 21

International Society on MCDM ...... 23

Past Conferences ...... 23

Sponsors ...... 25

Supporting Companies ...... 28

Scientific Program ...... 29

Author Index ...... 173

List of Chairs ...... 181

List of Participants ...... 183

Session Information ...... 196

Welcome Messages

Message from Dr. Murat Köksalan President, International Society on MCDM

Dear participants of the 25th International Conference on MCDM, Together with the organizers, I am delighted to welcome you to MCDM 2019. The first conference in this series was organized in 1975 in Jouy-en-Josas, France, by Herve Thiriez and Stanley Zionts. Since then, we have met roughly once every two years all over the world. The first MCDM conference I attended was in Cleveland, Ohio, USA in 1984. I have only missed a single conference since then. I have been fortunate to observe how our society and conferences grew and how our wonderful traditions developed. We have maintained an international composition of both our membership and the attendees to our conferences. Our society is among the most internationalized scientific communities with over 2700 members from about 100 different countries. We make an effort to maintain diversity in our committees as well. Participants of the Istanbul conference represent some 40 different countries. Those who are attending the MCDM conference for the first time will join many of us who have been enjoying the traditions of our society over many years. We have high-quality scientific programs and we keep the number of parallel sessions small to allow participants to follow important recent developments of interest to them. We make the conferences affordable to all and provide further support to those who are short of funding. Both the organizers and the society have been sensitive to the needs of students. Our society is among the very few scientific communities that have no dues and we are determined to keep it that way. The conference venues have alternated between continents. Organizers have included tastes of

1 local culture in conference activities. We have an outing to a location reflecting local heritage at every conference. Those of you who have been to our conferences know that there is a lot more than the great scientific programs at these events. I encourage you to enjoy all opportunities the organizers have prepared for you and more. In addition to the scientific program and material, the conference provides all lunches, coffee breaks, a couple of dinners, and the legendary banquet. These are great opportunities to spend time with fellow participants, to get together with old friends, and to make new friends. Many of us attest to the fact that long-lasting friendships and collaborations start at these conferences. As is the tradition, we will have the regular business meeting of our society on Thursday June 20, 2019 at 13:10. We will provide brief information about the society, discuss business, and have an open forum about the future. I would like to encourage all participants to attend and contribute to discussions. If you are not a member yet, I encourage you to apply at http://www.mcdmsociety.org as soon as possible. This is the second time we are meeting in Turkey. I had the privilege of chairing the organizing committee in year 2000 in Ankara. I know how much effort goes into organizing one of these conferences. There are many hard-working volunteers. I have been in close contact with the organizers of the Istanbul conference and I have observed their efforts to put together high-quality scientific and social programs. This is a great service to our community and I would like to thank the General Chair İlker Topcu and his colleagues for all their efforts. I look forward to meeting you and experiencing our traditions together in the fascinating city of Istanbul during June 17-21!

Murat Köksalan President, International Society on MCDM [email protected]

2

Message from Dr. İlker Topcu Conference General Chair, MCDM 2019

İstanbul’a hoşgeldiniz! I am very honored to host the 25th International Conference on Multiple Criteria Decision Making (MCDM 2019) at Istanbul Technical University, Istanbul, Turkey from June 16th to June 21st, 2019. This honor does not just belong to me; I have been encouraged and assisted by the greatest group of colleagues and friends anyone ever had to organize such a distinguished meeting consisting of a high-quality scientific program and an attractive social program. We, the organizing team from five different universities, are all very pleased and proud to welcome you to MCDM 2019. Istanbul Technical University is one of the leading universities of Turkey with a history stretching back to 1773. Istanbul Technical University Macka Campus, the conference venue, is in the heart of Istanbul and located very near the Bosporus strait through which the city's heart runs. Istanbul, the meeting point of cultures and civilizations and the capital of empires, is the only city in the world straddling both Europe and Asia. With historic streets, architectural masterpieces of old empires, exciting art scenes, panoramic vistas, and mystical beauties; Istanbul, the Capital of Culture, is truly a city that never sleeps. On the other hand, Turkey offers some of the world’s most important archaeological sites and historical monuments, vast and varied activities, welcoming Turkish hospitality and a delicious and assorted national cuisine. The official web site of the conference is www.MCDM2019.org. To check for news, updates, and recent developments, you may also follow MCDM 2019 on Twitter (twitter.com/mcdm2019) and on Instagram (www.instagram.com/mcdm2019). “MCDM beyond the Information Age” is the main theme of the conference. The submissions, covering several different topics and application areas support the theme strongly. We believe the scientific program is very rich and will be inspiring academicians and students as well as professionals in years to come.

3

Based on the traditions of the International Society on Multiple Criteria Decision Making, at MCDM 2019, a number of conference participants are financially supported by subsidized registration fees. For this conference, the International Society on MCDM and Creative Decisions Foundation deserve special recognition since they have offered grants for students and young researchers from developing and emerging countries. We have 260 participants coming from 5 continents and 38 countries. Among all participants, 84 of them are students who are the future of our Society. In the scientific program, 211 presentations are scheduled in mostly 5 parallel sessions. We have 38 contributed, 12 invited, and 3 special sessions along with 2 tutorials. We also have 4 plenary talks; a doctoral dissertation award finalists’ session as well as International Society on MCDM awardees’ talks. We all look forward to finding out about award recipients during our traditional banquette dinner. We worked hard to organize great social events to make the conference even more memorable. The “Welcome Reception” will be at Havuzbasi in Istanbul Technical University Macka Campus; the “Outing on Wednesday” will be at “Sakip Sabanci Museum” and “Emirgan Park”, followed by a “dinner cruise on the Bosphorus”; the “Conference Banquet” will be at “Sait Halim Pasa Mansion”, overlooking the seafront of Bosphorus. We hope you will enjoy the social program and find it worthy of the 25th organization of MCDM conferences! I owe a great deal of thanks to my friends General Co-chairs Özgür Kabak and Özay Özaydın as well as Organizing Committee Chair Şule Önsel Ekici and Program Committee Chair Serpil Sayın. I wish to thank Conference Secretary Çigdem Kadaifçi and Organizing Committee Members Sezi Çevik Onar, Bilal Ervural, Sait Gül, Gül İmamoğlu, Berk Küçükaltan, Başar Öztayşi, Seda Yanık, and Özgür Yanmaz. I would like to finish my message by expressing my gratitude to our supporting companies: Professional Conference Organizer Cormep and Event Management Company Beyaz Event as well as our sponsors: TÜBİTAK (The Scientific and Technological Research Council of Turkey), International Society on MCDM, Creative Decisions Foundation, Sabancı University Sakıp Sabancı Museum,

4

Çimtaş Pipe Fabrication and Trading Ltd. Co., Borçelik, Arama Participatory Management Consulting, Mavi, Anadolu Efes, Coca Cola İçecek, Dimes, Muratbey Peynir, Varta, Istanbul Metropolitan Municipality, Springer and Turkish Airlines.

Best wishes, Y. İlker Topcu MCDM 2019 Conference General Chair

5

General Information

Please bookmark the official web site of the conference www.MCDM2019.org for detailed information. To check for news, updates, and recent developments, you may also follow the meeting: • on Twitter (twitter.com/mcdm2019) and • on Instagram (www.instagram.com/mcdm2019) Venue

Istanbul Technical University, one of the leading universities of Turkey with a history stretching back to 1773, provides education in a modern educational setting and has a strong academic staff. The Macka campus which houses Faculty of Management, School of Foreign Languages and Conservatory is in the heart of Istanbul and located very near the Bosphorus strait. Address: ITU Management Faculty (Istanbul Teknik Universitesi Isletme Fakultesi), Macka St., 34367, Istanbul/TURKEY How to reach campus

ITU Management Faculty is located in one of the central districts of Istanbul. Welcome Reception, Opening Ceremony and Recommended Hotels are within walking distance of the conference venue. From Taksim Square, you may walk to Gondola lift station Taskisla and take the lift to Macka. The campus is just across the street from Macka station. From Besiktas, you may take the bus 30M Besiktas – Mecidiyekoy and get off at “Macka” stop. The stop is right in front of the campus. If you like walking up hills, you may also walk for 15 mins (1km) from Besiktas to Macka via Akaretler Yokusu (Suleyman Seba Street).

6

Getting around in Istanbul

The easiest way to get around in Istanbul is by taxi unless there is too much traffic. The fares are relatively cheap if we consider with distances. You should pay a minimum 10 TL to the driver due to minimum fare rule. In case of your journey fare above 10 TL, you will pay the fare that wrote on the taximeter. Make sure you have some Turkish liras with you. Most of the taxi drivers won't accept credit cards. We advise you to only take official taxis those have a sign on the roof with the word “taksi” on it. Furthermore, their taxi registration number (and sometimes company logo) should be clearly visible on the front doors and the roof. Most importantly, all official taxis have digital meters. You may hail a taxi on the street by waving your hand in the air. You can take a yellow taxi (cheapest one), a turquoise taxi (it is a bit expensive), or a black taxi (most expensive and very luxury). You may use the mobile applications BiTaksi or Itaksi to call a taxi online. Via these apps, you can pay with your credit card or cash. Public transport in Istanbul comprises a bus network, various rail systems including underground ones, funiculars, and maritime services. Participants will be given a public transport map with details. Registration Desk

The registration desk is located on the first floor of ITU Management Faculty. We recommend picking up your conference material as soon as you arrive on Sunday before the Welcome Reception. You will require your badge and ticket for the Reception. The registration desk will be open at: • 15:00 – 19:00 on June 16, Sunday • 08:00 – 17:00 on June 17, Monday • 08:00 – 17:00 on June 18, Tuesday • 08:00 – 14:00 on June 19, Wednesday • 08:00 – 15:30 on June 20, Thursday • 09:00 – 14:00 on June 21, Friday

7

Registration Fee

Registered participants will receive a badge and tickets for social events giving them access to the conference venue as well as the conference materials. Participants are requested to wear their badge visibly at all times. As usual, regular or student registration covers attendance at all sessions and conference materials as well as admissions to “Welcome Reception on Sunday”, “Outing on Wednesday”, “Conference Banquet on Thursday”, Lunches, and Coffee Breaks. Accompanying person registration covers the same except the admission to sessions and conference materials. Internet Access

You may connect to wireless internet via “eduroam” using your university credentials. Also, the participants will be provided a temporary “ITU/NET Misafir” account to access internet. The passwords will be given in the conference package. Lunch and Coffee Breaks

Lunches are included in the registration fee and will be served from Monday to Friday in Ari Kovani. Tickets are required for lunches. Coffee breaks will be at the second floor of Management Faculty. Important Phone Numbers

Emergency Phone Numbers in Turkey: • Police: 155 • Ambulance: 112 • Fire: 110 All three numbers are toll-free and do not require roaming plans.

8

Disclaimer

The organizing committee is not liable on any damage on health or property of any conference participant or accompanying person in the course of or resulting from participation in the conference. Participants are advised to have their own insurance. Authors are responsible for the content and copyright of their work.

9

MCDM 2019

Committees

Chairs

• İlker Topcu (General Chair, Istanbul Technical University)

• Özgür Kabak (General Co-chair, Istanbul Technical University)

• Özay Özaydın (General Co-chair, Doğuş University)

• Şule Önsel Ekici (Organizing Committee Chair, Doğuş University)

• Serpil Sayın (Program Committee Chair, Koç University) Organizing Committee

• Sezi Çevik Onar (Istanbul Technical University)

• Bilal Ervural (Istanbul Technical University)

• Sait Gül (Beykent University)

• Gül İmamoğlu (İstanbul Technical University)

• Özgür Kabak (Istanbul Technical University)

• Çigdem Kadaifçi (Doğuş University)

• Berk Küçükaltan (Trakya University)

• Şule Önsel Ekici (Doğuş University)

• Özay Özaydın (Doğuş University)

• Başar Öztayşi (Istanbul Technical University)

• İlker Topcu (Istanbul Technical University)

• Seda Yanık (Istanbul Technical University)

• Özgür Yanmaz (Istanbul Technical University) Scientific Program Committee

• Emel Aktaş (Cranfield University) • Umut Asan (Istanbul Technical University) • Adiel Teixeira de Almeida (Federal University of Pernambuco) • Sarah Ben Amor (University of Ottawa) • Matthias Ehrgott (Lancaster University)

10

• Gabriele Eichfelder (TU Ilmenau) • José Rui Figueira (Technical University of Lisbon) • Salvatore Greco (University of Catania) • Zahir Irani (University of Bradford) • Özgür Kabak (Istanbul Technical University) • Birsen Karpak (Youngstown State University) • Kathrin Klamroth (Bergische Universität Wuppertal) • Murat Köksalan (Middle East Technical University) • Banu Lokman (Middle East Technical University) • Kaisa Miettinen (University of Jyväskylä) • Enrique Mu (Carlow University) • Özay Özaydın (Doğuş University) • Serpil Sayın (Koç University) • Hsu-Shih Shih (Tamkang University) • Johannes Siebert (University of Bayreuth) • Seyhan Sipahi Nişel (Istanbul University) • Theo Stewart (University of Cape Town) • İlker Topcu (Istanbul Technical University) • Füsun Ülengin (Sabancı University) • Jyrki Wallenius (Aalto University) • Margaret Wiecek (Clemson University) • Constantin Zopounidis (Technical University of Crete) Awards

International Society on MCDM Awards Since 1992, the International Society on Multiple Criteria Decision Making presents, at each of its international conferences, up to three awards to individuals in honor of their outstanding, long-lasting, and influential contributions to the field of MCDM.

11

The awards will be presented at the 25th MCDM 2019 Conference in Istanbul, Turkey. The awardees will be announced at the conference banquet on June 20, Thursday. Their talks will be held on June 21, Friday. These awards are: The MCDM Gold Medal This is the highest honor that the International Society on Multiple Criteria Decision Making bestows upon a scholar who, over a distinguished career, has markedly contributed to the theory, methodology, practice and professional development of MCDM. The MCDM Edgeworth-Pareto Award This is the highest distinction that the International Society on Multiple Criteria Decision Making bestows upon a researcher or practitioner of MCDM who has demonstrated a high level of creativity in developing novel areas of application of MCDM and associated methodology, markedly influencing the form of MCDM practice. The Georg Cantor Award This is the highest form of recognition that the International Society on Multiple Criteria Decision Making bestows upon a researcher who has personified the spirit of independent inquiry in developing innovative ideas in the theory and methodology of MCDM, significantly expanding the tools available to MCDM practice. MCDM Doctoral Dissertation Award 2019 The International Society on Multiple Criteria Decision Making bestows the MCDM Doctoral Dissertation Award upon a scholar who has recently obtained a doctoral degree and demonstrated excellence in her or his doctoral dissertation research in Multiple Criteria Decision Making or a related field. The MCDM Doctoral Dissertation Competition identifies and recognizes outstanding doctoral dissertation research, completed on January 1, 2017 or later, in the development of theory, methodology, and/or the application of theory or methodology to MCDM. The finalists for this year are:

12

• Fritz Bökler, Osnabrück University, Germany “Output-sensitive Complexity of Multiobjective Combinatorial Optimization Problems with an Application to the Multiobjective Shortest Path Problem”

• Mohammad Ghaderi, Pompeu Fabra University, Spain “Preference Disaggregation: Towards an Integrated Framework”

• Andreia Guerreiro, University of Coimbra, Portugal “Portfolio Selection in Evolutionary Algorithms” Their talks will be given on June 19, Wednesday from 1:10pm to 2:25pm in Doctoral Dissertation Award Session at Mustafa Kemal Amfisi. The final decision will be made after the talks and the result will be presented at the conference banquet. Plenary Sessions

Plenaries from Monday to Thursday will be held in Mustafa Kemal Amfisi. Plenaries on Friday will be held in ITU Management Faculty. Opening Session Opening session will be held on June 17, Monday between 09:00 – 09:40. Invited Plenary Talks The following four distinguished speakers have agreed to give a talk at MCDM 2019. June 17, Monday Pınar Keskinocak, Georgia Tech, USA 09:40 – 10:40 “Quantitative models for decision-support in healthcare applications” June 18, Tuesday Kathrin Klamroth, Univ. of Wuppertal, Germany 11:10 – 12:10 “Multi-Objective Combinatorial Optimization – Beyond the Bi-objective Case” June 19, Wednesday Tuomas Sandholm, Carnegie Mellon, USA 11:10 – 12:10 “New Results for Solving Imperfect-Information Games” June 20, Thursday Luis Vargas, University of Pittsburgh, USA 11:10 – 12:10 “Voting with Intensity of Preferences”

13

Doctoral Dissertation Award Session Doctoral dissertation award session will be held on June 19, Wednesday between 13:10 – 14:25. Business Meeting Business meeting will be held on June 20, Thursday between 13:10 – 14:10. Society Award Talks Society Award talks will be given on June 21, Friday between 09:30 – 12:00. Closing Session Closing session will be held on June 21, Friday between 14:40 – 15:10. Tutorial Sessions

There will be two tutorial sessions. First one will be given by Pekka J Korhonen, Aalto University School of Business titled “MOLP vs. DEA – Relatives or Friends” on June 18, Tuesday between 16:00 – 17:40. The second tutorial will be given by Orrin Cooper, University of Memphis titled “Improve Your Decisions by Learning and Experiencing the ANP Best Practices on June 17, Monday between 13:50 – 17:40. Orrin Cooper’s tutorial is also a special session sponsored by CDF and a tribute to Thomas L. Saaty & a great opportunity to learn ANP. Parallel Sessions

Contributed Sessions There are 151 talks assigned to 38 contributed sessions in the MCDM 2019 scientific program. They cover the following topics:

• Advanced Optimization Techniques • Advances in MCDA • Advances in MCDM Theory • AHP/ANP

14

• Contemporary Approaches in MCDM • Data Envelopment Analysis • Evolutionary Algorithms • Fuzzy sets and approaches • Group Decision Making • Interactive Multi-Objective Optimization • Multi-Objective Metaheuristics • Multi-Objective Optimization • Multiple Criteria Decision Aiding • Outranking Methods • Preferences, Comparisons and Uncertainties • Ranking or Sorting Invited Sessions Additionally, there are 48 talks assigned to 12 invited sessions in the scientific program. They cover the following topics:

• Building MCDM/A Models: Practical and Methodological Issues • Continuous Multiobjective Optimization with Engineering Applications • Data Science meets Multiple Criteria Decision Making • Hybrid Approaches in MCDM • MCDM in Solving Urban Problems using Big Data • MCDM/A for Assessment Issues in Healthcare and LCA • MCDM/A Models with Flexible and Interactive Preference Modeling • Multiobjective Optimization Applications • Multiobjective Route Planning • New Approaches in MCDM • Solution Methods for Various Multiobjective Optimization Problems • Theory and Applications of ANP Special Sessions Finally, there are 10 talks assigned to 3 special sessions:

• AHP in Practice as ‘’Decision Conference’’ in Turkey - sponsored by ARAMA • AHP/ANP Applications in Industry 4.0 - sponsored by ÇİMTAŞ • AHP/ANP Applications in Production and Manufacturing - sponsored by BORCELIK

15

Session Identifiers

Every session is identified with a 5-character code which means the following: • The first three characters correspond to the day of the week when the session takes place: MON, TUE, WED, THU, FRI. • The fourth character displays the time slot of the day. For instance, WED-1 means that all sessions with this identifier are assigned to the first time slot of Wednesday. • The fifth character represents the type of the session as follows: o P corresponds to plenary talks o A, B, C, D, E, and F display different parallel streams including contributed, invited, and special sessions o O stands for opening session o M stands for International Society on MCDM award talks o U stands for business meeting o L stands for closing session Instructions for Session Chairs

Each session chair will be given a session follow-up and evaluation form with necessary information for the session. The volunteer responsible for the room will pick the form at the end of the session. In case of no-show presentations, Session Chairs are kindly requested to pause the session for the duration of the no-show presentation to allow cross-session participations. Instructions for Presenters

Each presentation is given 20 minutes. An additional 5 minutes are allocated for questions and discussion at the end of 20 minutes. The slides should be in English. Please, bring your presentation on a USB memory stick and save it as a PDF or PowerPoint document. Find your room and meet the Session Chair before your oral presentation. Also make sure that you upload your slides on the computer before the session starts.

16

Publications

Springer Book Chapter Selected abstracts that are presented at the conference will be invited to be considered as chapters in the book titled “Multiple Criteria Decision Making Beyond the Information Age: Selected works from MCDM 2019 Istanbul” within the Springer‘s Book Series Contributions to Management Science. Editors: Y. İlker TOPCU, Ozay OZAYDIN, Ozgur KABAK, and Sule ONSEL EKICI Journal of Enterprise Information Management Special Issue Abstracts that are presented at the conference will have the opportunity to be considered as articles in the Special Issue of Journal of Enterprise Information Management, a Social Science Citation indexed journal. Special issue title would be “Contemporary Strategic Decision Making through Multiple Criteria Approaches in the New Industrial Era” Guest editors: Emel AKTAS, Berk KUCUKALTAN, and Y. İlker TOPCU International Journal of the Analytic Hierarchy Process Special Issue Abstracts that are presented at the conference will have the opportunity to be considered as articles in the Special Issue of International Journal of the Analytic Hierarchy Process (IJAHP), a SCOPUS-indexed journal, that publishes papers about research and applications of the Analytic Hierarchy Process (AHP) and Analytic Network Process (ANP). Guest Editors: Y. İlker TOPCU and Ozay OZAYDIN Journal of Global Optimization Special Issue Abstracts that are presented at the conference will have the opportunity to be considered as articles in the Special Issue of Journal of Global Optimization, a Science Citation indexed journal, that deals with theoretical and computational aspects of seeking

17 global optima and their applications in science, management and engineering. Guest Editors: Kaisa Miettinen and Serpil Sayın Journal of Multi-Criteria Decision Analysis Special Issue Abstracts that are presented at the conference will have the opportunity to be considered as articles in the Special Issue of Journal of Multi-Criteria Decision Analysis, a SCOPUS-indexed journal, that provides an international forum for the presentation and discussion of all aspects of research, application and evaluation of multi-criteria decision analysis, and publishes material from a variety of disciplines and all schools of thought. Guest Editors: Thierry Marchant and Marc Pirlot Annals of Operations Research Special Issue Abstracts that are presented at the conference will have the opportunity to be considered as articles in the Special Issue of Annals of Operations Research, a Science Citation indexed journal, that deals with key aspects of operations research, including theory, practice, and computation. Guest Editors: TBA Optimization and Engineering Special Issue Abstracts that are presented at the conference will have the opportunity to be considered as articles in the Special Issue of Optimization and Engineering, a Science Citation indexed journal, that promotes the advancement of optimization methods and the innovative application of optimization in engineering. Guest Editors: Jussi Hakanen and Richard Allmendinger

18

Social Program

Welcome Reception Welcome reception will be at Havuzbasi in Istanbul Technical University Macka Campus on June 16, Sunday from 18:00 to 20:00. Group Photo The group photo will be shot at the steps inside the building in front of the entrance of Mustafa Kemal Amfisi on June 18, Tuesday between 12:20 to 12:50. Outing The “Outing on Wednesday” will be on June 19, Wednesday at Sakip Sabanci Museum and Emirgan Park, followed by a dinner cruise on the Bosphorus. Sabancı University's Sakıp Sabancı Museum, a private fine arts museum, is located at one of Istanbul's oldest settlements on the Bosphorus. Sakıp Sabancı Museum presents a versatile museological environment with its rich permanent collection, the comprehensive temporary exhibitions that it hosts, its conservation units. Emirgan Park, tulip’s homeland, is one of the biggest parks in Istanbul. Emirgan Park includes approximately ninety species of tree and three thousand varieties tulips. It is said to be the most beautiful park in Istanbul. Boat ride on the Bosphorus, a splendid panorama, is a unique experience for seeing the beauties including palaces, monuments, and mansions accompanied by trees and flowers along the European and Asian sides of the Bosphorus. Buses for the first section of the Outing will depart at 14:40 from in front of ITU Management Faculty. There will be an English speaking guide at the museum. Due to capacity constraints, half of the participants will start at the museum, while the other half will be taken to Emirgan Park. Later the halves will switch places. First group visit of the museum will be between 16:00 – 17:00, while second group will visit the museum between 17:30 – 18:30. After visiting both the museum and the park, the participants will be taken to the Emirgan quay for boarding the boat at 19:00. The boat will disembark the participants at Karaköy around 23:00.

19

Banquet The conference banquet will be at Sait Halim Pasa Mansion on June 20, Thursday. The Mansion, located at the heart of the Bosphorus and overlooking the seafront, was built in 1878. Its decorative style is a mélange of French and Egyptian architecture. 19th-century architectural elegance meets with 21st-century service quality. Mesmerizing Bosphorus view will be the backdrop to our banquet. The night will begin with a cocktail in the Mansion at 18:45. It will be followed by a dinner in the Garden at 20:00. The busses will depart from in front of ITU Management Faculty at 17:45, they will depart from the Mansion around midnight to return to the same place to drop-off participants.

20

MCDM 2019 in Numbers

There are four plenary talk sessions; one doctoral dissertation award finalists’ session, and one International Society on MCDM awardees’ talk session. There are 211 talks scheduled in five or six parallel sessions. 38 of these parallel sessions are contributed, 12 of them are invited, while 3 or them are sponsored special sessions. 260 participants are coming from 5 continents and 38 countries. As expected, the largest number of participants are from Turkey with 86 participants, followed by United States with 23 participants. Germany and United Kingdom have 12 participants. The number of participants from Brazil and Algeria is 11.

Continents

21

Countries

86

23

12 12

11 11

9

8 8 8

7

6 6

5

4 4

3 3 3 3 3

2 2 2 2 2 2 2 2

1 1 1 1 1 1 1 1 1

IRAN

INDIA

OMAN

CHILE

SPAIN

JAPAN

I T A L Y

ISRAEL

BRAZIL

SERBIA

RUSSIA

TAIWAN

TURKEY

FRANCE

GREECE

CANADA

TUNISIA

POLAND

NIGERIA

NORWAY

ALGERIA

SWEDEN

FINLAND

BELGIUM

ROMANIA

HUNGARY

GERMANY

SLOVENIA

PAKISTAN

COLOMBIA

PORTUGAL

P . R . C H I N A

SOUTH AFRICA

NETHERLANDS

UNITED STATES

R.N. MACEDONIA CZECH REPUBLIC UNITED KINGDOM

Connecting Children with Nature MCDM 2019 Organizing Committee is aware of the carbon footprint caused by the participants for their attendande to MCDM 2019. To even partially compensate for our effect, we made donations in names of all participants to TEMA Foundation - The Turkish Foundation for Combating Soil Erosion, for Reforestation and the Protection of Natural Habitats (www.tema.org.tr).

22

International Society on MCDM

MCDM can be defined as the study of methods and procedures by which concerns about multiple conflicting criteria can be formally incorporated into the management planning process. The purposes of the Society are to develop, test, evaluate and apply methodologies for solving multiple criteria decision making problems, to foster interaction and research in the scientific field of multiple criteria decision making, and to cooperate with other organizations in the study of management from a quantitative perspective. These purposes are to be carried out by:

• facilitating communications, • providing academic assistance with ideas, methods and contacts whenever and wherever needed, • sponsoring international conferences for the dissemination of results, • encouraging special purpose workshops and student exchanges, • publishing the electronic newsletter and • encouraging an environment of international colleagueship and friendship across the globe. Past Conferences

24th International Conference: July 10-14, 2017, Ottawa, Canada, Sarah Ben Amor. 23rd International Conference: August 3-7, 2015, Hamburg, Germany, Martin J. Geiger. 22nd International Conference: June 17-21, 2013, Málaga, Spain, Francisco Ruiz. 21st International Conference: June 13-17, 2011, Jyväskylä, Finland, Kaisa Miettinen. 20th International Conference: June 21-26, 2009, Chengdu, China, Yong Shi, Shouyang Wang. 19th International Conference: 2008, New Zealand, Matthias Ehrgott. 18th International Conference: 2006, Chania, Crete, Greece, Constantin Zopounidis.

23

17th International Conference: 2004, Whistler, British Columbia, Canada, Bill Wedley. First MCDM Winter Conference (16th International Conference): 2002, Semmering, Austria, Mikulas Luptacik, Rudolf Vetschera. 15th International Conference: 2000, Ankara, Turkey, Murat Köksalan. 14th International Conference: 1998, Charlottesville, Virginia, USA, Yacov Y. Haimes. 13th International Conference: 1997, Cape Town, South Africa, Theo Stewart. 12th International Conference: 1995, Hagen, Germany, Günter Fandel, Thomas Gal. 11th International Conference: 1994, Coimbra, Portugal, Joao Climaco. 10th International Conference: 1992, Taipei, Taiwan, G.H. Tzeng, P.-L. Yu. 9th International Conference: 1990, Fairfax, USA, Richard Soland, Ambrose Goicoechea. 8th International Conference: 1988, Manchester, UK, A. Geoff Lockett, Gerd Islei. 7th International Conference: 1986, Kyoto, Japan, H. Nakayama, Y. Sawaragi. 6th International Conference: 1984, Cleveland, Ohio, USA, Yacov Y. Haimes. 5th International Conference: 1982, Mons, Belgium, Pierre Hansen. 4th International Conference: 1980, Newark, Delaware, USA, Joel Morse. 3rd International Conference: 1979, Knigswinter, Germany, G. Fandel, T. Gal. 2nd International Conference: 1977, Buffalo, New York, USA, Stanley Zionts. 1st International Conference: 1975, Jouy-en-Josas, France, Herve Thiriez, Stanley Zionts.

24

Sponsors

TÜBİTAK (The Scientific and Technological Research Council of Turkey) supports MCDM 2019 with 2223-B Program. Support for Students & Young Researchers

The Creative Decisions Foundation was established in 1996 by Thomas L. Saaty and his wife Rozann Whitaker Saaty to promote the cause of rational decision making and spread awareness of the Analytic Hierarchy Process (AHP) theory of decision making created by Dr. Saaty.

International Society on MCDM aims to develop, test, evaluate and apply methodologies for solving multiple MCDM criteria decision making problems, to foster interaction and research in the scientific field of multiple criteria decision making, and to cooperate with other organizations in the study of management from a quantitative perspective. Outing Sponsor

Sabancı University Sakıp Sabancı Museum presents a versatile museological environment with its rich permanent collection, the comprehensive temporary exhibitions that it hosts, its conservation units, model educational programs and the various concerts, conferences and seminars held there.

25

Session Sponsors

Çimtaş Pipe Fabrication and Trading Ltd. Co., founded in 2002, is a market-leading company providing engineering, procurement, manufacturing, welded fabrication, machining, and painting services of LSAW pipes, fittings and pipe spools for the power, process, OEM piping systems and pressurized equipment to power, oil, gas and chemicals, water, mining and metals industries with a global vision to be the “best in class”. The strengths of the company are: focus on safety, excellence in services, world class quality, on time delivery, shortest cycle times and continuous improvement with Lean Production Approach.

Borçelik founded as the first private and second largest flat steelmaker of Turkey in 1990. Borçelik started its activities in 1994 to produce “cold-rolled steel sheet coils”. Borçelik continues its activities under the partnership of Borusan Holding and ArcelorMittal, one of the largest global steelmakers. The Company has a total production capacity of 1.5 million tons with the Borçelik brand and a metal processing capacity of 500,000 tons with the Kerim Çelik brand, making it Turkey’s biggest galvanized steelmaker, offering the highest quality. The company’s strong market position is based on dynamic labor force, innovative approach, continuous investment for self-development and growth, customer- oriented service and quality approach.

ARAMA is a participatory management consulting company that pioneers participative methodologies in the renewal, restructuring and reconfiguring of organizations, institutions, industries, regions and other social systems including international and national bodies. The company was established in Istanbul and Ankara in 1995, and the Pennsylvania office was established in 1998. Since then, about 1000 participative methods have been applied in Turkey. Today ARAMA is

26 the only consulting company in Turkey that focuses only on the development and application of participative methods and processes. Food, Beverage & Product Sponsors

Incorporated in Istanbul in 1991, today Mavi is recognized as a highly successful, global lifestyle brand. With strong denim roots, Mavi has evolved into an international apparel and accessories brand.

Anadolu Efes is the industry leader with 3 breweries, 2 malt factories, and 1 hops processing plant in Turkey. They are an international company that operates in Russia, Georgia, Kazakhstan, Ukraine, and Moldova in addition to Turkey. Their products are consumed in more than 80 countries across the world. They are Europe’s 6th and World’s 14th largest beer producer in terms of sales volume.

CCI is a multinational beverage company which operates in Turkey, Pakistan, Kazakhstan, Azerbaijan, Kyrgyzstan, Turkmenistan, Jordan, Iraq, Syria and Tajikistan. As one of the key bottlers of the Coca-Cola system, CCI produces, distributes and sells sparkling and still beverages of The Coca-Cola Company.

DİMES is founded in Tokat, the heart of Anatolia in the 1960s and it is the first local fruit juice brand in Turkey. DİMES is the biggest fruit juice producer with its 3 factories and the leading brand of the juice category in households in Turkey. They are exporting their products to over 100 different countries. DİMES is one of the leading and well-known fruit juice brand in Turkey.

Muratbey is a registered trade mark in approximately 60 countries. Patent of the original and similar forms of our innovative cheese Helix, which is created cleverly and of which its design is

27 registered in approximately 60 countries, belongs to us. We are exporting our own trademarks to approximately 50 countries including all EU, US, Japan, Middle East, Northern African Countries, Balkans, TRNC, Turkic Republics.

VARTA is a key international manufacturer of consumer batteries. The company is the market leader in Germany, Austria and Switzerland, as well as in other European countries. VARTA Consumer identifies new market developments and the customers; needs early on and caters to them with new and innovative products.

Istanbul Metropolitan Municipality (IMM) is the administrative center of Istanbul, one of the world’s most populous cities with a total population of around 15 million residents in its metropolitan area. IMM provides with emergency and social services, libraries and museums, investment and financial services, licensing and transportation services.

Springer is a leading global scientific, technical and medical portfolio, providing researchers in academia, scientific institutions and corporate R&D departments with quality content through innovative information, products and services. Other Sponsors International Association of Turkish and Ukrainian Businessmen (TUİD) is a businessmen association founded in 2004 by Turkish businessmen in Ukraine and with the auspices of the embassy of Republic of Turkey in Kiev. Supporting Companies

Cormep is the Professional Conference Organizer for MCDM 2019. Beyaz Event is the event planner for MCDM 2019.

28

Scientific Program

1. On Rationality Conditions for MONDAY Multi-Attribute Choice Behavior Pekka Korhonen, Aalto University School of Business, Finland, MON-0-O [email protected] Opening Session Jyrki Wallenius, Aalto University, Finland, [email protected] Monday 9:00-9:40 - Room: Mustafa Kemal Peng Xu, Aalto University School of Amfisi Business, Finland, [email protected] Chair: Ilker Topcu Tolga Genc, Marmara University, MON-1-P Turkey, [email protected] Plenary Session 1 This paper deals with rationality conditions for choice behavior. It bears similarity to Monday 9:40-10:40 - Room: Mustafa Kemal Samuelson’s Revealed Preference Theory, Amfisi although it deviates from it in significant ways. We explore two different types of Chair: Murat Köksalan choice settings: (1) win-win setting, where Quantitative Models for Decision- consumers consider which goods they want Support in Healthcare Applications to add to their basket, (2) tradeoff setting, where consumers make pairwise choices *Pınar Keskinocak, Georgia Tech, between different (efficient) baskets, where United States, [email protected] they have to give up in some goods to gain in other goods. We study the decision-maker’s Health systems focus on broad set of rationality conditions in both settings. The activities including prevention, screening, perspective is that of an outside observer, diagnosis, and treatment. Most health who can use the conditions to judge, whether systems face multiple objectives and limited the choices are rational or not. The basket is resources while striving to provide high filled either with different or identical goods. quality care and promote health and well- The key underlying theoretical assumptions being. Data-driven quantitative models, in our paper are increasing and concave embedded into decision-support systems, single dimensional utility (value) functions have a significant potential for improving with decreasing marginal values (win-win patient and population outcomes, efficiency, setting) and the Kahneman-Tversky Prospect and effectiveness in healthcare. In this Theory model of choice with loss aversion presentation we will discuss a few (tradeoff setting). We use an empirical applications from organ transplant, experiment to motivate and verify our vaccination, screening, and workforce considerations. allocation decisions. 2. An Efficient Interactive Approach MON-2-A to Ordinal Pairwise Comparisons Contributed Session: Advances in *Elena Moshkovich, University of MCDM Theory - Behavioral and Montevallo, United States, Computational Issues [email protected] Monday 11:10-12:50 - Room: Galata Tower Alexander Mechitov, University of Chair: Jafar Rezaei Montevallo, United States, [email protected]

Many multiple criteria decision methods rely on a decision maker’s preferences to resolve

29 conflicts between different criteria and being Matteo Brunelli, University of Trento, able to rank order alternatives. Rather often Italy, [email protected] a decision maker’s preferences are elicited in the form of pairwise comparisons of In this research, we explore the uncertain objects. Some methods, like AHP or M- input problem in Best-Worst Method (BWM), Macbeth elicit this information in a cardinal a promising pairwise comparison Multi- form, other methods require elicitation of Criteria Decision-Making (MCDM) method preferences in an ordinal form, e.g., ZAPROS which has received abundant attention by or ARACE. Research shows that people make the research community since its proposal. mistakes in their judgments leading to By using BWM, the weights of criteria can be inconsistency of preferences which results in obtained by solving the linear or nonlinear an unstable solution to the problem. The program based on the paired comparison question of eliciting consistent information is preferences provided by decision-makers. of upmost importance to the success of the Compared with Analytic Hierarchy Process decision aids so we need methods for (AHP), BWM needs fewer comparison data discovering, reducing, and/or eliminating but can generate more consistent inconsistency in the pairwise comparisons comparisons and produce more reliable even if the information is elicited in an results according to previous analyses. In ordinal form. Methods of Verbal Decision order to tackle the uncertain preferences Analysis such as ZAPROS and ARACE require provided by decision-makers, the belief the decision maker to fill in comparison structure is introduced to BWM, a concept of matrices of a rather substantial sizes (e.g., preference degree adopted by evidential 12 x12 or even larger). Excel-based system reasoning approach. With their flexible ORCON helps find inconsistencies in such distribution assessments, belief structures matrices and shows efficient steps in can effectively capture both precise data and resolving contradictions in preferences various types of uncertainties such as through re-visiting some comparisons by the ignorance and ambiguity in subjective decision maker. At the same time a decision judgments. A new approach is proposed to maker has to carry out many pairwise enable BWM to cope with this form of comparisons and then do additional information, in which the belief degrees of evaluations in the process of elimination of the given preferences are taken into inconsistencies. We propose an interactive consideration. In real-world contexts, it is approach to the process of pairwise likely that group decision-making be comparisons by a decision maker. preferred over individual decisions. In fact, Transitivity of preferences is used to many of the decision processes that are minimize the number of comparisons carried made in the real world occur in group out by a decision maker. The procedure settings to make the decision-making more though does not provide enough information comprehensive and reasonable. However, for verification of the validity of the results. the uncertainty contained in the estimations The authors introduce the principle of provided by the group decision-makers and “double check” and propose an algorithm for the inconsistency involved in the pairwise necessary additional comparisons to meet comparisons can and often lead to unreliable this principle. Efficiency of the proposed and unstable results. So it is necessary to approach is evaluated through a computer measure the uncertainty and inconsistency modeling and is illustrated through several degree of the given preferences to decide examples. ORCON-I Excel-based system may whether the result can be adopted or their be used to carry out the process. evaluation should be turned back to the decision-makers to modify. In this research, 3. Group Best Worst Method Based a new consistency ratio measurement is on Belief Structure proposed to examine the inconsistency degree of decision-makers in group *Fuqi Liang, Delft University of decisions. Compared to the original Technology, Netherlands, f.liang- consistency measurement proposed in the [email protected] original BWM, this new consistency measure Jafar Rezaei, Delft University of is easier to compute and has clearer Technology, Netherlands, algebraic meaning, interpretation and [email protected] properties. To check the uncertainty degree

30 of decision-makers, we adopt the total solution, it is also important to check the uncertainty measure of belief structures robustness of the results, which is the main proposed by Pal and Bezdek et al. in 1993, aim of this study. Using analytical approach, which has low computational complexity and we find the maximum theoretical sum of the satisfies a set of desirable axioms. The interval ranges of the optimal weights. This combination of inconsistency degree and the theoretical value differs for problems with uncertainty degree of beliefs is then used to different number of criteria. The value is then reflect the reliability of the decision-makers. used in a ratio to calculate the robustness Based on the weights of criteria obtained index of a problem which has been solved by from the BWM and the reliability degrees of the non-linear BWM. The proposed decision-makers, the assessments for robustness index, along with the consistency alternatives provided by decision-makers are ratio of the model, provides enhanced aggregated by Evidential Reasoning insights into the reliability of the results. The algorithm, which is particularly useful in findings show that (i) there is a high dealing with belief structure information, and correlation between the consistency ratio of the alternatives are finally rated and the pairwise comparisons provided by the selected. We conclude that incorporating the decision-maker and the robustness index of belief structure into group BWM enables the final results; (ii) pairwise comparisons decision-makers handle their beliefs in a with the worst consistency ratio, i.e. one, flexible and versatile way, and combining might be associated with extreme low and evidential reasoning approach and BWM high values of robustness index, i.e. zero or together considers the effect of belief one. degrees on decision-makers’ consistency Rezaei, J. (2015). Best-worst multi-criteria and the advantage of Dempster’s rule, making this method a good option for decision-making method. Omega, 53, 49-57. decision aiding. MON-2-B 4. Best Worst Method: Robustness Invited Session: Theory and Applications Analysis of ANP *Jafar Rezaei, Delft University of Monday 11:10-12:50 - Room: Maiden's Technology, Netherlands, Tower [email protected] Chair: Orrin Cooper Best Worst Method (BWM) (Rezaei, 2015) is 1. The Development of a Conference a pairwise comparison-based multi-criteria decision-making method that offers a City Selection Model as a Decision structured way to make the comparisons. Making Managerial Support Tool After identifying a set of relevant decision- Using AHP/ANP Best Practices making criteria by the decision-maker, the best (e.g. the most important) and the worst Enrique Mu, Carlow University, United (e.g. the least important) criteria are States, [email protected] specified by the decision-maker. Pairwise *Milagros Pereyra-Rojas, University of comparisons are conducted between the two Pittsburgh, United States, reference criteria (best and worst) and the [email protected] other criteria using a nine-point scale. The Orrin Cooper, University of Memphis, two input pairwise comparison vectors (best- United States, [email protected] to-others and others-to-worst) are used in a Michael Peasley, Middle Tennessee minmax optimization model to find the State University, United States, optimal weights of the criteria. The original BWM formulates the problem as a non-linear [email protected] model, which may result in multiple optimal The goal of the present study is the weights (optimal weight intervals). The development of an actionable model that reliability of the weights obtained from the can be used, in particular by executive method is checked using the consistency directors and managers of professional ratio, which is calculated after solving the associations, as a managerial decision optimization model. Since the non-linear support tool for conference city selection. model may provide more than one optimal While this topic has been sparsely addressed 31 in the literature, it has been typically CRITERIA) 1 City Logistics Business Best discussed as a specific one-time event or as Practices, Hotel Room Costs, Meeting an academic study (Mu, 2006). This study Venues, Policy Best Practices) 2 Travel will develop a usable conference city Considerations (Airfare Costs, Visa Costs, selection decision support tool using a Other Accommodations, City Appeal, practical decision making approach (Mu and Personal Safety) 3 Academic & City Financial Pereyra-Rojas, 2018a) and recently & Political Support (Academic Alignment, proposed AHP/ANP best practices (Mu, Financial Support, Political Support) Our Cooper and Peasley 2017, 2018). This study study contributions are as follows: • From a is positioned within the stream of research to professional point of view, we are addressing translate AHP/ANP studies into actionable the problem of selecting a conference city-a managerial decision-making frameworks common but important decision in (Mu and Pereyra-Rojas, 2018; Mu and professional associations. • Furthermore, Pereyra-Rojas, 2018b, 2019) and will be we are providing a simple to use everyday illustrated with a case study in a professional decision tool rather than approaching it as a association with long experience using AHP. one-time academic problem. • From a Our first step has been meeting with the CEO methodological point of view, we will be and her top three directors to review and reporting this study using the recently update an ANP model developed to predict developed best practice checklist developed conference attendance in 2009 (Mu, 2006). for AHP/ANP models (Mu, Cooper and The results, at the time, predicted the Peasley, 2018). conference would be a record success in terms of attendance and this was 2. ANP to Ascertain Cooperative and corroborated in the field. Unfortunately, it Competitive Capabilities in the was the only time it was used since the Context of Coopetition model was designed to estimate participation for the 2009 meeting. Since it Ramesh Dangol, Youngstown State has been almost ten years since this study, University, United States, the top directors in the organization (only one [email protected] of them was present in the first study) *Birsen Karpak, Youngstown State considered it was time to review what had University, United States, been learned since that time as well as to [email protected] develop a new model that could be used as a decision support tool that could be used The cooperative and competitive routinely. See model below (Table 1). Table (coopetition) literature shows that firms can 1 – Participation Prediction ANP Model innovate and gain a competitive advantage CLUSTER (NODES) 1 Alternatives (City 1, City by cooperating with competitors For 2, City 3) 2 Risks (Public Safety) 3 example, Gnyawali & Park (2011) show that Political/Strategic Factors (Political Stability, fierce rivals Sony and Samsung cooperated Economic Factors, Academic Alignment) 4 to create new LCD screens and, Costs (Airfare, Visa, Lodging, Living Costs 5 consequently, increased their market share Travel Factors (Travel Time, Time of the Year, in the television market. Although the Language, Touristic Appeal) Our next step coopetition literature shows the benefits of has been meeting with the different directors cooperating with competitors, what is separately to elicit their criteria for ignored in the literature is the extent to which conference site selection as well as the firms can benefit from such relationships. associated weighting. Consistency was We argue that the extent to which firms can addressed with each director when needed. benefit from coopetition depends on the The different priorities were aggregated relative differences in cooperating through a multiplicative process following competitor’s cooperative and competitive standard procedure (Mu and Pererya-Rojas, capabilities. When cooperating competitors 2018a). Again, the goal was to develop a city possess different levels of cooperative selection model that could be used capabilities, they would not be able to recurrently as a decision support tool for identify and acquire resources embedded in their annual meeting city selection. The each other’s operating routines essential for model is shown below. Table 2 – Conference innovation. As a result, they would not be City Selection AHP Model CRITERIA (SUB- able to take a full advantage of cooperative

32 relationships. In contrast to cooperative Competitors with similar levels of capabilities, firms rely on competitive cooperative capabilities would be viewed as capabilities to appropriate benefits created better alternatives compared to firms with by cooperating with competitors. Imbalances dissimilar levels of the same capabilities. In in competitive capabilities could allow a firm contrast, transaction cost risks (R) are likely with high levels of competitive capabilities to to be high when firms have dissimilar levels appropriate the lion’s share of benefits at the of competitive capabilities. Transaction cost expense of its cooperating competitors. If risks equal “0” when firm are similar in their imbalances in cooperative and competitive competitive capabilities. This paper shows capabilities can limit cooperating that firms can use ANP to prioritize among competitors’ ability to innovate and create competitors based on the marginal benefits the potentials for one firm to appropriate they can generate by cooperating and the more benefits at the expense of the others, risks imposed by differences in competitive then we need develop methods that firms capabilities. can use to accurately determine their capabilities and those of their competitors. 3. Centralized vs Decentralized Unfortunately, we are not aware of any Warehousing Configuration: An research that shows how firms can Analytic Network Process Approach accurately determine their own and their competitors’ capabilities levels. Therefore, *Ozden Bayazit, CWU, United States, the primary goal of our study is to develop a [email protected] method based on ANP that firms can employ Warehousing plays a vital role in providing to evaluate capabilities and, subsequently, the desired level of customer service at the form an enduring cooperative relationships. lowest possible total cost and became a Our ANP model will take four elements into competitive resource for many companies. account when selecting a competitor with While some companies send their inventory whom a firm wishes to establish a to multiple warehouses to stay competitive cooperative relationship – marginal benefits and provide the desired level of customer (B), opportunity costs (O), expenses (E) and service, the other companies choose to keep transaction cost risks (R). Marginal benefits it in one centralized location. The most (B) is the excess benefits firms create by common warehousing configurations are cooperating with competitors, whereas either centralized or decentralized. As the opportunity costs (O) is the marginal benefits number of warehouses a company uses firms would have created if they had formed increases, the system becomes more a relationship with a different partner. decentralized. On the other hand, if a Therefore, the net marginal benefit a firm company chooses to keep its entire inventory realize is simply marginal benefits minus in one single, central warehouse, the system opportunity cost (B- O). Expenses (E) is becomes centralized. In centralized amount of resources a firm devotes to warehousing, operation costs will be lower cooperative relationship. Finally, transaction and customer service levels will increase; cost risk (R) is the potential costs resulting however shipping costs will be much higher the likelihood that cooperating competitors when compared to a decentralized decide to use resources created by configuration. Very clearly, both systems cooperating in areas not covered by the possess advantages and disadvantages for contract and R will work to magnify E. the companies. Even knowing the pros and Transaction cost risks vary from 0-1; risk cons doesn’t necessarily make the decision equals 0 when competitors perceive no risk. an easy one. Choosing the most optimal Therefore, the total cost of forming warehousing configuration decision involves relationship with a competitor can be both quantitative and qualitative factors and expressed by E+R. The cost benefit ratio can represents a multi-criteria decision-making be calculated by dividing B-O by E+R (or (B- problem. Due to interdependencies among O)/(E+R)), which is the basis for selecting a the criteria that affect the companies’ competitor from a choice set for establishing decision on choosing the most optimal a cooperative relationship. Priorities of warehousing configuration, we have decided alternatives (different competitors) on B is to use the Analytic Network Process (ANP) to determined by whether or not firms have approach the problem. The ANP is a similar levels of cooperative capabilities. methodology for multi-criteria problems 33 where there are feedback and pairwise comparisons, the decision maker interdependence among decision attributes will recognize that in the original and alternatives. The ANP is a generalization comparisons an error was made; and believe of the Analytic Hierarchy Process (AHP) which the updated comparisons are not only more is represented by a network. A network has consistent but also make more sense and clusters of elements, with the elements in are a better representation of what they are one cluster being connected to elements in measuring. Often times the revision process another cluster or the same cluster. Our helps the decision maker clarify important network included four clusters namely, points or definitions of the criterion and how "Alternatives", "Costs", "Customer Service" it was being measured which improves the and "Other". Each cluster had different quality of the overall decision. In summary, criteria. For example "Costs" cluster included by trying to improve the data quality by inbound transportation costs, outbound checking for consistency, not only can the transportation costs, inventory holding costs, consistency index be improved but it can also warehouse operating costs, and cost of rush provide significant insights to the decision delivery factors. Comparisons are made to maker leading to revisions beyond new evaluate both configuration alternatives with pairwise comparisons that are simply more respect to many factors inside the clusters. consistent. The Linking Coherency Index The priorities derived from pairwise (LCI) is an innovative method to test for comparison matrices are entered as parts of coherency in ANP Supermatrices. Coherent the columns of the unweighted supermatrix. data can be defined as self-consistent and The final outcome presented interesting non-contradictory with respect to a particular results. The proposed model is one of the system. Coherency can also be thought of as first to apply the ANP in choosing the most a “super consistency test” or a test for optimal warehousing configuration. The consistency at the “system” level of the major limitation of the technique is that it entire Supermatrix. By performing data was applied for an illustrative example of a quality tests at the level of the Supermatrix logistics company in the U.S. However, the additional relationships and data are results will provide guidance to companies available to test for data quality than what is that are trying to decide whether available only at the level of the pairwise decentralization or centralization is best for comparison matrix. The consistency index them as the decision is a strategic one for can test for data quality among comparisons many competing in today’s global markets. across a single unit of measurement. Coherency uses data from the comparisons 4. Coherency: Improving Data Quality across multiple units of measurement and Reducing Comparisons in the combined with the relationships between the ANP different units of measurement that were provided by the decision maker to test for *Orrin Cooper, University of Memphis, data quality. Coherency allows the decision United States, [email protected] maker to test the data both within a single Idil Yavuz, Dokuz Eylul University, unit of measurement and the relationships Turkey, [email protected] between the different units of measurement. Linking Estimates (LE) are an important When developing the AHP, Saaty recognized component used when calculating the LCI. the value of research that would improve the The LE can also be used to provide another quality of decision data. The Linking advantage to decision makers by potentially Coherency Index (LCI) is a data quality check reducing the number of comparisons that are for ANP decisions. Engaging in data quality required in ANP decisions. After reviewing checks can not only increase the statistical the LCI, the value of this new data quality likelihood of making a good decision; but it check, testing the coherency of the also provides an opportunity for important Supermatrix, and the value using the LE to reflections about the decision and priorities potentially reduce the number of pairwise that have been provided. Many decision comparisons will be demonstrated through a makers have made pairwise comparisons, neat interactive example. tested for consistency at the level of the pairwise comparison matrix, and needed to revise the comparisons. When revising the

34

MON-2-C in a study area of Mleta, regrouping 74 land units, in the western part of Algeria. Contributed Session: MCDM for Group Decisions 2. Implementation of Cumulative Belief Degree Approach to Group Monday 11:10-12:50 - Room: Dolmabahçe Palace Decision-Making Problems Under Hesitancy Chair: Özge Şahin Zorluoğlu *Nurullah Güleç, Ankara Yıldırım 1. Methodological Approach for a Beyazıt University, Turkey, Group Decision-Making in Land [email protected] Suitability for Agriculture Özgür Kabak, Istanbul Technical *Mendas Abdelkader, Centre des University, Turkey, [email protected] Techniques Spatiales, Algeria, In multi-criteria group decision-making [email protected] problems (GDM), decision-makers may have Mebrek Abdellah, Geomatic, Algeria, hesitancy in their evaluations on assigning [email protected] linguistic terms. This hesitancy may Sadouni Zakaria, Geomatic, Algeria, negatively affect the quality of the solution. [email protected] Recently, the use of the Hesitant Fuzzy Linguistic Terms (HFLTs) has increased Mekranfar Zohra, Geomatic, Algeria, rapidly due to the flexibility it provides to [email protected] decision-makers. Using HFLTs, decision- This research presents a methodological makers can assign more than one linguistic approach to perform a spatial multi-criteria terms in their evaluations. There are a lot of decision analysis based on the view of studies in the literature for solving GDM multiple experts or stakeholders and the problems with HFLTs. However, when HFLTs importance of each criterion. It takes into and other evaluation formats such as direct account the impact of the policy makers. value assignment, classical fuzzy sets, Twelve criteria (Water reserve easily linguistic terms etc. are used in the same utilizable, Drainage, Permeability, Potential problem, the methods in the literature are of hydrogen (pH), Electrical Conductivity, limited. The Cumulative Belief Degree (CBD) Active limestone, Cation Exchange Capacity, approach, based on fuzzy linguistic terms Soil texture, Soil useful depth, Slope, and belief structure, is a solution method Availability of labour, Proximity (roads)) applied in many complex multi-criteria GDM grouped in five factors (agronomic, planning, problems under different assessment land enhancement and improvement, soil methods. In this study, a multi-criteria GDM conservation and environment protection, problem in which the evaluations are and socio-economic conditions) will be used provided with different evaluations formats in this study. Criteria will be evaluated with including HFLTs is considered. A method the Analytic Hierarchy Process (AHP) by based on CBD approach has been experts in the subject area. The obtained developed. Specifically, transformation weights will be introduced into a Consensus formula for converting HFLTs to CBDs is Convergence Model (CCM) to get consensual proposed. The proposed method has been criteria weights which will be used to assess applied to a sales manager selection the land suitability for agriculture. Individual problem in literature. The method has been preferences of policy makers which include shown to be convenient for GDM problems other factors that cannot be presented as with HFLTs. maps or spatial data will be simulated in order to allow involved policy makers to rank 3. A Stochastic Multi-Actor Multi- the land units of the study area using a Criteria Decision Analysis Approach preferential method. All the required to Evaluate Alternative Marine Fuels processing methods will be implemented into a GIS environment. The methodological *Dina Aspen, NTNU, Norway, developments are motivated by an [email protected] application on land suitability for agriculture Børge Andreas Johansen, NTNU, Norway, [email protected]

35

Selection of alternative marine fuels and uncertain, it is possible to distinguish power systems is a recurring decision preferable alternatives for each stakeholder problem for Norwegian ferry operators and group thanks to the SMAA-component of the concession authorities. When ferry crossings approach. When these results are combined are tendered, perspectives of multiple using equal weighting of stakeholder groups, stakeholder and decision-maker groups play following the MAMCA method, we see that it into the overall credibility and acceptance of is still possible to identify promising the final decision. Deployment of new and technology options by evaluating stochastic partly unproven technologies may also imply dominance based on robustness measures. highly uncertain outcomes, which further This proves the point that even with diverging adds to the complexity of the decision opinions, it may be possible to identify problem. We present an approach to tackle alternatives that are satisfactory from an both these issues in ferry concession overall stakeholder perspective. The problems by combining principles from multi- composite approach may be a useful tool in actor multi-criteria decision analysis public procurement or other large (MAMCA) with stochastic multicriteria investment problems where outcomes are acceptability analysis (SMAA). The multi- uncertain and the perspectives of multiple actor perspective allows to analyze stakeholder groups needs attention. This alternatives from the view of stakeholder includes problems where options must be groups both separately and in concert to qualified or disqualified prior to a tender call distinguish promising solutions. To ensure and problems where comparative satisfactory handling of data, the approach performance needs to be evaluated once also allows to utilize different multi-criteria offers are submitted. decision making (MCDM) aggregation procedures. By modeling performance and 4. Hierarchical Group Decision preference information as stochastic Making Approach for Project parameters, rank robustness of alternative Portfolio Selection technology options are used to evaluate their overall performance. To illustrate the SMAA- *Özge Şahin Zorluoğlu, Istanbul MAMCA approach, a Norwegian ferry Technical University, Turkey, concession case study in a World Heritage [email protected] Area is explored from the perspective of Özgür Kabak, Istanbul Technical operators, concession authorities, and the University, Turkey, [email protected] local community. Due to the different nature of criteria sets for various actor groups, Hierarchical Group Decision Making (HGDM) different MCDM procedures are utilized to is a decision-making problem with a large ensure adequate handling of data. These number of decision makers (DMs) or experts encompass a value function method organized in a hierarchical structure. In this (TOPSIS), an outranking method (ORESTE) problem, DMs not only propose alternatives and a tender specific scoring model used for but also provide evaluations of alternatives ferry concession evaluations by Norwegian with respect to their own criteria. The Public Road Authorities. Nine fuel technology objective of this study is to define a novel options are explored, including biofuels, HGDM problem, a new perspective in group fossil fuels, electricity, hydrogen and hybrid decision making. The characteristics of variants. Performance of alternatives across HGDM are described in four dimensions criteria for the three stakeholder groups are which are based on the position and number defined as stochastic parameters. of stakeholders (DMs, experts, etc.), Preferences are also modeled as uncertain, alternatives and criteria. The stakeholders assuming complete lack of information on are divided into hierarchical units. Some the relative importance of criteria for each stakeholders may be located out of the stakeholder group. The results from the case hierarchy. A large number of people are study shows that different fuel technology involved in the process. Each hierarchical alternatives exhibits first order stochastic unit has its own alternatives and criteria. dominance from the view of different There are also common criteria that are stakeholder groups. This illustrates the point evaluated by experts out of hierarchy. that even if preference information is Furthermore, for this problem, a practical missing, and performance information is solution approach which is easy to

36 understand and apply by each stakeholder is AHP in Practice as ‘’Decision proposed. HGDM problem involves a large Conference’’ in Turkey number of DMs, alternatives and a few criteria sets. Therefore, it has a complex *Oğuz Babüroğlu, Sabancı University structure. Some difficulties may be and Arama Consulting, encountered while explaining the whole [email protected] structure to the stakeholders (DMs, experts, Derya Köker, Alexion Corporation, problem owner, analysts, etc.). That’s why, a [email protected] convenient solution methodology is proposed for the problem that is articulate This symposium is designed to offer the for stakeholders. A solution methodology practice side of an age old multi criteria based on pairwise comparisons and ratings decision making methodology; analytical that are functional and frequently used multi hierarchy process (AHP). AHP has been criteria decision-making approaches are applied in more than 100 situations ranging utilized for the newly proposed HGDM from the choices Galatasaray Sports Club approach. One of the significant faced in the 2005, the future of the Topkapi contributions of the study is the application Palace, to solve the organ transplantation of the HGDM structure to Project portfolio bottleneck in İstanbul, many challenges selection (PPS) problem. PPS is a process to confronting different municipalities, industry obtain the best portfolio of projects for an organizations and regional development organization under consideration of resource agencies, UNDP Turkey strategic plan, for scarcity. Many stakeholders, who are associations’ priorities and to navigate classified in a hierarchical manner, propose between the strategic projects and choices projects and evaluate projects with respect confronting many different Turkish and to various sets of criteria. The aim is to form international corporations operating in a portfolio of projects that will be performed. Turkey. AHP in practice takes the form of PPS problem is modeled time in the literature ‘’Decision Conference’’ which is a form of in various ways by considering one or more conference that is designed to enable the objective functions, resources or criteria, participation of the relevant stakeholders to labor force, etc. However, most of the studies the setting and to the projects to be just consider the top management’s evaluated and prioritised. We have found preferences not the opinions of DMs or that it is much better to genrate the model experts who propose these projects and and the the projects through another others who will perform them. This could lead participatory methodology called ‘’search to bias in the selection process. It is conference’’. Tje decision conference important to consider preferences of all typically follows the search conference and a levels within the hierarchy. Therefore, it committment conference follows the seems quite appropriate to apply the HGDM decision conference. The triangulation of a approach to the PPS problem. Properties of variety of conferencing methodologies the HGDM are discussed based on the facilitates the participation of large numbers application. It is seen that HGDM approach of stakeholders and a logical strategy improves efficiency with fewer questions. formation and decision making track to More accurate results are achieved with the engage them. We will go over one such participation of DMs from different levels of process of collaboration to address the organ the organization. transplantation problem situation in İstanbul. The collaboration process that MON-2-D embodied search, decision and Special Session: AHP in Practice as committment conferences resulted in the ‘’Decision Conference’’ in Turkey - increase of transplantation cases from 500 to 5000 in a period of a couple of years. sponsored by ARAMA Together with the sponsor of the whole Monday 11:10-12:50 - Room: Topkapı process Mrs Derya Köker, the marketing Palace manager of a different company at the time, we will discuss the trials and tribulations as Chair: Özay Özaydın well as the outcomes of the real life case.

37

targets in isolation as additional inputs we preferred gradually expanding (or chaining) MON-2-E the input space. Considering a target, its order of joining the input space is a point to Invited Session: Data Science meets pay attention. Undoubtedly, the optimal Multiple Criteria Decision Making configuration can be revealed via exhaustive Monday 11:10-12:50 - Room: Hagia Sophia search. However, we do not want either generalization performance degradation or Chair: Jussi Hakanen process overhead. To cope with these issues, 1. Benefiting the Target Relations in we create a routine that screens the best learned target and prioritizes it. We used a Supervised Learning: Multiobjective byproduct of the tree ensemble classifiers, Extensions vs Space Expansion that is out-of-bag-errors, that does not Methods require any extra computational effort or data. In addition, it helps to capture the *Esra Adıyeke, Boğaziçi University, nonlinear relations between targets. We set Turkey, [email protected] up a series of experiments to test out our Mustafa Baydoğan, Boğaziçi University, methods both in terms of time complexity Turkey, [email protected] and predictive performance. We compared the proposed methods with their state of the In our study we consider supervised learning art. Initially, we compared the multiobjective problems for data sets with multiple outputs methods for further analyses. We carried out (targets). The targets may exhibit scale another series of experiments to test the differences and a combination of different proposed methods against different and scales require either data preprocessing or competitive chaining strategies. The results model modifications to keep the learning suggest that, our chaining strategy process avoid from being dominated by a outperforms the multiobjective methods and specific target. Data preprocessing is the difference between their predictive cricitized as it may result errors in performance is statistically significant. measurements and may change the Moreover, though the predictive statistical properties of the data. Hence, we performances of our chaining method and its aim to develop classifiers so that they are state of the art are not statistically insensitive to scale of the data. In addition, significant, still our chaining method yields we further aim to augment the prediction the best rank. In addition, we empirically quality of the classifiers by exploiting the showed that theoretical time complexity of target relations. Derivation of alternative our method is consistent with the ways towards the given concerns, is the calculations. We performed the theme of this study. The first strategy adapts aforementioned tests on a regression an existing single objective multitask setting. In other words, all the targets are learning (MTL) model to its multiobjective homogeneously continuous in the MTL version. The objectives consist of experiments. However, the targets of a data indicators of learning quality, i.e. mean set may exhibit difference in terms of scales. squared error. We changed the standard To clarify, a mixture of categorical and multitask design of tree classifiers that numerical targets is a typical instance of considers the quality criterion as an equally such case. We did an additional experiment weighted sum form. We proposed several by using an artificial data set with one tree derivation heuristics so that we can take categorical and one numerical target. advantage of approximated Pareto optimal Clearly, the targets do not agree in terms of solutions. The second strategy benefits from scale. We compared the predictive the idea of transforming the learning performances of the selected multiobjective problem. Noting that, our study is in a method and our chaining method with a multiple target setting, and in the former weighted sum policy. The results suggest strategy targets take place in the learning that, our multiobjective design performs process merely as elements of output space. slightly better than the rest. As a second strategy, we transform the problem so that targets have dual roles as inputs and outputs. In order to treat the

38

2. A Grid-based Algorithm to good classifiers for difficult imbalanced Generate Well-Dispersed classification problems and a comprehensive treatment of distinct positive Nondominated Solutions for a Three and negative error levels may lead to Objective SVM Formulation for the important classification performance Imbalanced Data Classification improvements. References Aşkan, A., & Problem Sayın, S. (2014). SVM classification for imbalanced data sets using a multiobjective *Serpil Sayın, Koç University, Turkey, optimization framework. Annals of [email protected] Operations Research, 216(1), 191-203. Gökhan Kirlik, University of Maryland Medical System, United States, 3. Application of the Choquet [email protected] Integral in K-Means Clustering Method Classification problems in which instances that belong to one class outnumber the *Kerbouı Roumeissa, USTHB, Algeria, other class of instances significantly is [email protected] referred to as the imbalanced data Abbas Moncef, USTHB, Algeria, classification problem. These imbalanced [email protected] data sets are commonly encountered in real- life problems; however, performance of well- Preference models often represent the known classifiers is limited in such cases. overall degree of utility of an alternative in Various solution approaches have been terms of the aggregation of several degrees proposed for the class imbalance problem of local utility, each of which relates to a using either data-level or algorithm-level specific criterion. Methods for learning modifications. We build on a three objective preference models from observed L1-norm SVM formulation that was preference data have focused mainly on suggested by Aşkan and Sayın (2014). Their adjusting the aggregation function while key idea is to treat the sum of empirical assuming the local utility functions that need errors for the two classes separately in the to be given. This article presents the Choquet SVM formulation. Their solution integral as a mathematical tool for learning methodology is based on reducing the three classification. Although it is widely used as a objective problem into biobjective ones by flexible aggregation function in areas such as parameterization of one of the error sums. At multi-criteria decision making, the Choquet the chosen level of a parameter value for one integral is much less known in machine type of empirical error, the remaining learning so far. The Choquet integral bicriteria problem is solved for its entire combines monotony and flexibility in a Pareto frontier. This leads to cross-sections healthy and elegant mathematical way, but of the original three-dimensional Pareto set also has additional features that make it in the outcome space. The classifiers attractive from a machine learning point of obtained in this way are evaluated on a view. For example, it provides measures to validation set and a best performer is picked. quantify the importance of individual In this study, we propose obtaining a predictors and the interaction between representation of the three dimensional groups of criteria, which promotes the nondominated set directly, without using the interpretability of a model. In the same cross-sectional grid structure. We concept, we can interpret the Choquet experiment with a representation that is integral in a multi-criteria distance from obtained by using a grid of both error sums clustering algorithms based on the most simultaneously and solving a subproblem delegated partitions, namely k-means. When based on the achievement scalarizing this aggregation function takes into account function. The proposed approach does not the interaction between criteria without provide quality guarantees on the losing the information of the criteria. In representation that is delivered and it relies addition, the proposed approach is rather on the ability of the achievement scalarizing modest in multi-criteria aggregation, function to target a specific region in the compared to the classical Euclidean distance outcome set. Numerical experiments that makes the combination with the k- demonstrate that the approach can deliver means aggregation and the famous Choquet

39

Integral aggregation function. A generalised measurements from the point of view voice data has been illustrated in the problem quality and intelligibility. which is compared with the usual k-means algorithm. The results of this practical MON-2-F approach were compared and found to be Contributed Session: MCDM for Project more accurate, easier to understand and, Selection most importantly, require less time to process. Monday 11:10-12:50 - Room: Basilica Cistern 4. New Approach to Speech Signal Enhancement via Multicriteria Chair: Jose Rui Figueira Optimization and Adapted 1. A Multi-Objective Mathematical Metaheuristic Model of the Multi-Period *Said Ouznadji, USTHB, Algeria, Substitutable Projects Location- [email protected] Allocation Problem Chaabane Djamal, USTHB, Algeria, *Mohamed Essalah Salah, University of [email protected] Sfax, Tunisia, [email protected] Messaoud Thameri, Ecole supérieure Ali Amira Ghorbel, University of Sfax, Chabati, Algeria, Tunisia, [email protected] [email protected] Arij Kilani, University of Sfax, Tunisia, This paper proposes a new contribution in [email protected] the field of speech signal enhancement. This Younes Boujelbene, University of Sfax, is the development of a method combining a Tunisia, multi-criteria approach to optimization with a [email protected] suitable metaheuristic. Indeed, there are several methods dedicated to the Today, the formulation of some improvement of the speech signal which is combinatorial optimization problems often corrupted by a noise. The classical revolves on the combination of several spectral subtraction method introduced by existing models. In this paper we formulate a Boll causes a distortion of the signal and multi-objective model of the multi-period gives rise to an annoying musical noise in the substitutable projects location-allocation ear. To solve the problem Berouti introduces problem. This problem combines three sub- two parameters alpha and beta, of control problems which are the multi-period that it fixes experimentally. Also, for us, it is a substitutable resources allocation problem, question of finding a compromise between the two dimensional bin-packing problem distortion and musical noise by resorting to a and the plant location-allocation problem. In biobjective optimization method combined order to treat the formulated multi-objective with a metaheuristic of the type optimization model, we present the goal programming by swarm of particles. The implementation of approach and we obtain a goal programming this study aims to compare the performance model. we propose a general procedure to of the algorithm developed with the classical solve the obtained goal programming model. methods enhancement speech signal 2. Integrating Case-Based Analysis corrupted by additive noise present in hostile environments in a single channel context. and Fuzzy Optimization for Selecting Objective and subjective evaluation of the Project Risk Response Actions proposed method is performed using quality *Yao Zhang, Northeastern University, and intelligibility measures (SNR, PESQ ...). The performance of the algorithm is studied China, [email protected] in the Airport, Babble, Car…, noise This article proposes a method based on environments and the results obtained show case-based analysis and fuzzy optimization that the improvement is better, even superior to provide decision support in project risk to the classical algorithms of spectral response. The main steps of the method are: subtraction developed by Boll and Berouti, 1) the formulation of alternative risk with a clear improvement of the response actions (RRAs) based on case- based analysis, and 2) the determination of

40 the optimal set of RRAs using a fuzzy part based on its Expected Transition Impact optimization model. Based on the method, (ETI) that signals the strength of its expected project managers (PMs) can find out development outcomes. Over the past 3 alternative RRAs and further determine an years, the institution has been developing a optimal set of RRAs. Compared with the methodology and an IT application to score existing RRA selection methods, the method the ETI of projects based on a coherent set proposed in this study makes two of criteria. The presentation will present the contributions. First, the case-based method methodology and link it with the MCDA (Multi- and the optimization method are integrated Criteria Decision Analysis) framework to for decision support in project risk response. discuss the core decisions made. In doing so The alternative RRAs can be obtained by the principal scientific contribution will be (i) using the case-based method; while the to provide an applied example of a decision- optimal set of RRAs is further selected out making methodology and its application and from the alternatives with the optimization (ii) to bridge the gap between theory and model. Second, the fuzzy set theory is practice by discussing to what extent MCDA applied to evaluate the risk probability, risk principles can be directly applied in a large impact and the similarity between risks in the organisations with a legacy of adhoc, expert- RRA selection process. The advantage of based decision-making. Specifically, we will using the fuzzy set theory is that the PMs and explain how economists assigned scores to experts can make the evaluations with different criteria, the structure of the score linguistic terms, which is more suitable for aggregation, how we approach concepts human perception in actual situations. And such as independence of preferences, the quality of the adopted method of compensation, or uncertainty, and how IT measuring the fuzzy distance is over most of systems may support or limit efficient the existing methods through ambiguity and decision-making. An interesting feature of fuzziness which are two main attributes of the tool is that it combines decision-making fuzzy numbers. Some managerial suggestion criteria and a set of incentives for bankers to and implication are drawn from the results of increase development outcomes. Moreover, the article. First, to perform better risk it is embedded in the EBRD’s wider project response in future, it is suggested that cycle by linking decision-making to organizations should always capture a long- monitoring. We connect criteria (or project term perspective, with an awareness of objectives) to indicators that will be tracked keeping documents of all handled historical throughout the life of the project. In doing so projects. Second, since each RRA obtained we hope to improve future decisions by from alternative historical cases needs to be having data on the delivery of past projects adapted in accordance with the current (feedback loops). There may be different situations, adaptation costs should also be ways to enter information about a project in considered when allocating budget for the application and this may lead to different selecting RRAs. scores. Moreover, a committee reviews project submissions on a weekly basis and 3. Assessing Development Impact to has the possibility to amend scores. Hence Guide Investment Decisions – the ETI (development outcomes score) is the Lessons from Implementing TOMS at result of the application/methodology, the EBRD decisions taken by the Banker in entering the project in the application, and the review by *Pawel Krasny, EBRD, United Kingdom, economists. We will discuss how project [email protected] scoring is embedded in a wider institutional setting and is a result of a compromise The European Bank for Reconstruction and between sound decision-making, internal Development (EBRD) is an International constraints and legacy. We will also present Financial Institution (IFI) supporting private issues in implementing this approach across sector development and market economies the EBRD. Development outcomes in Europe, the Middle East North Africa assessment systems are increasingly used in (MENA) region, Central Asia and the development institutions but are often seen Caucasus. It does so mainly by providing as bespoke and not funding to projects that support its policy transferable/comparable tools. The goals. The decision to approve a project is in presentation will aim at setting a frame for a

41 discussion between such organisation and problems. More precisely, given the with specialists in MCDA and we will give a increasing interest of historical cities to few examples on how other IFIs approach the restore their cultural heritage the integrated assessment of development outcomes. multiple criteria decision aiding analytical tool proposed in this paper has an important 4. A Multiple Criteria Methodology potential of being used in the future. for Prioritizing and Selecting Portfolios of Urban Projects MON-3-A *José Rui Figueira, Technical University Contributed Session: Advances in of Lisbon, Portugal, MCDM Theory - Applications in Diverse [email protected] Industries Maria Barbati, University of Portsmouth, Monday 13:50-15:30 - Room: Galata Tower United Kingdom, Chair: Nitin Harale [email protected] Salvatore Greco, University of Catania, 1. A Fuzzy Analytic Network Process Italy, [email protected] Approach for Healthcare Service Alessio Ishizaka, University of Quality Evaluation by Using Portsmouth, United Kingdom, Neutrosophic Sets [email protected] Simona Panaro, University of *Hatice Camgöz Akdağ, Istanbul Portsmouth, United Kingdom, Technical University, Turkey, [email protected] [email protected] Kemal Konyalioglu, Istanbul Technical This paper presents an integrated University, Turkey, methodology supporting decisions in urban [email protected] planning. In particular, it deals with the prioritization and the selection of a portfolio Tuğçe Beldek, Istanbul Technical of projects related to buildings of some University, Turkey, [email protected] values for the cultural heritage in cities. In Service quality has always been debatable in particular, our methodology has been order to investigate how to improve and how validated to the historical center of Naples, to find deficiencies. Especially in healthcare Italy. Each project is assessed on the basis of sector, not only for quality improvement but a set of both quantitative and qualitative also patient satisfaction, service quality is criteria with the purpose to determine their perceived to be increased step by step but level of priority for further selection. This step which factors to be improved, renovated or was performed through the application of the taken into consideration are not always Electre Tri-nC method. This method is a obvious and not easy to decide. In the multiple criteria outranking based model for literature, there exist many decision-making ordinal classification (or sorting) problems methods that are being used to prioritize and allows to assign a priority level to each different alternatives. These methods are project as an analytical \recommendation" very important when a strategic decision will tool. A set of resources (namely budgetary be made at a big organization. In hospitals, constraints) as well as some logical seen as complex systems, there are many constraints related to urban policy different factors for the evaluation of service requirements have to be taken into quality, including internal and external consideration together with the priority of factors. Thus, it is necessary to prioritize projects in a portfolio analysis model which factors should be improved in order to permitting to identify the effi cient portfolios provide an efficient service for patients. In and to support the selection of the most this case, Multi Criteria Decision Making adequate set of projects to activate. The (MCDM) tools are very useful to prioritize process has been conducted thanks to the service quality. On the other hand, the interaction between analysts, municipality method of Analytic Network Process (ANP) is representative and experts. The proposed widely used in MCDM problems. Especially, methodology is generic enough to be applied even if the factors affecting service quality in in other territorial or urban planning hospitals, are seen as objective, Fuzzy ANP

42 can be considered as a very effective method 3. A Fuzzy Cognitive Mapping as the factors are generally evaluated under Approach to Turkish Football uncertainty. Thus, fuzzy MCDM methods are very suitable but depending on the set which Industry is used during MCDM process, the priorities Gorkem Altug, Doğuş University, Turkey, can change based on the method and fuzzy [email protected] sets. There exist differents sets which are Ozgur Yanmaz, Istanbul Technical used in fuzzy MCDM methods. In 1998, University, Turkey, [email protected] Smarandache has introduced neutrosophic sets as a general view of intuitionistic fuzzy *Cigdem Kadaifci, Doğuş University, sets, incorporating a new parameter in order Turkey, [email protected] to define the concepts of a membership and Football is attracting a considerable amount nonmembership. In this study, it is aimed to of people around the world. According to the put forward A Fuzzy Analytic Network Process Nielsen Sports survey in 2018, 43% of the Approach for Healthcare Service Quality 1.7 billion participants indicate that they are Evaluation by Using Neutrosophic Sets to either interested or very interested in prioritize service quality in hospitals. football. Being the top sports capturing the Neutrosophic sets will provide a different interest worldwide, it has a 75% share prioritization of service quality factors among Turkish participants. Besides the affecting especially patient satisfaction in popularity, football is a developing industry hospitals. Furthermore, it is aimed to define considering the financial results. The annual which factors should be improved in order to financial report of Deloitte demonstrates that reach the desired patient satisfaction. European football market size is €25.5 2. A Closed-Loop MCDM Method billion in 2017, where 58% is generated by “big five” European leagues (Premier Learning Model for Intelligent League, Bundesliga, La Liga, Serie A, and Manufacturing Ligue 1). Turkish Super League is the sixth *Jei-Zheng Wu, Soochow University, largest market in Europe with €734 million revenue. Under these social and financial Taiwan, [email protected] circumstances, football clubs are turning into To accomplish next-generation intelligent corporate companies. In order to improve the manufacturing, developing complete sets of quality of football, to maintain financial measurement and a framework is critical to stability and/or to increase their profitability, ensure effective multi-criterion decision- they have to adapt to the dynamic nature of making (MCDM) decisions. Research on the industry. Even international associations these topics is lacking. As a result, this study enforce football clubs to be more transparent aims to develop a closed-loop learning model and financially mature through the restrictive to implement decisions based on selected regulations, a systematic tool is needed to MCDM methods that are trained and support the strategic actions and sustain validated as effective methods. Through a both the sports and financial achievements. extensive literature review, this study To address this issue, a Fuzzy Cognitive summarizes well-known MCDM methods for Mapping (FCM) method is proposed to effectiveness comparisons. Various examine the objectives of football clubs in normalization and aggregation methods and Turkey. Eighteen objectives are determined parameters settings will be discussed. A based on a detailed literature review and illustrative case in the semiconductor expert opinions. The objectives can be manufacturing is conducted. The results of classified into three groups: administrative, this study demonstrate how selection financial, and sports. FCM allows to process of MCDM methods can become represent the causal relationships between automatic and intelligent by integrating novel the objectives and to examine the long term machine learning methods. behavior of the system based on the current circumstances. By using the method, a decision support tool to Turkish football industry is provided to build a strategic framework.

43

4. Supplier Selection in Data-Driven then encoded into numerical scale according Fashion Industry by Using MCDM to the applied MCDM method. The software packages authors used to conduct this study Methods include Python, R, MATLAB, and Excel *Nitin Harale, Ecole centrale de Lille, depending on the suitability and flexibility of France, [email protected] their libraries. Authors’ main objective in this paper is to identify the best performing In fashion supply chain management, MCDM method for the selected research selection of suppliers is a critical and integral problem considering the complexities it part of overall business processes. By involves. Authors expect that the results from selecting potential suppliers, who provide this study will provide well suited and high quality products to the customers at the accurate approach for dealing with high right time, fashion companies ensure their dimensionality of supplier selection criteria business growth, efficiency and profitability. and sub-criteria, and the pool of competing However, due to the ever changing customer suppliers in the market. It is also expected choices and the multitude of qualitative and that this study will address the conflicting quantitative criteria based on which fashion information generated in the big data retailers often select their suppliers, supplier environment of companies’ business models selection process entails highly complex while making group decisions for local as decision making. Moreover, due to the well as global sourcing. Authors envisage advent of e-commerce business models and that this paper will provide valuable insights advanced database technologies, it is and contribute to the enhanced group increasingly becoming challenging to select decision making in the fashion industry in the appropriate criteria among many regard to supplier selection. Moreover, it will conflicting, inconsistent, and contradictory provide a proper direction for the further ones based on which supplier selection can research in this area. be achieved. In this paper, authors aim to address this challenge by using MCDM MON-3-B methods, viz. Fuzzy-AHP, TOPSIS, Contributed Session: Consistency PROMETHEE, VIKOR methods. MCDM methods are popular for solving multi-criteria Issues in AHP and group decision making problems. The Monday 13:50-15:30 - Room: Maiden's research problem in this paper revolves Tower around the challenge of selecting the best possible suppliers considering the range of Chair: Jorge I Romero criteria and sub-criteria from both retailers’ 1. The Accuracy of the Consistency- and customers’ point of view. A questionnaire based survey approach is Compatibility Relationship in the AHP adopted to evaluate and identify highly Context important qualitative and quantitative *Adriana Miclea, Bucharest University criteria. The participants were supply chain of Economic Studies, Romania, managers from four European fashion companies who specialize in range of [email protected] customized fashion products. Given the An extensive review of the scientific literature increasing trend of big data technologies in on consistency and compatibility indexes in major industries, fashion retailers are now the context of ANP theory, together with able to gauge as to what predominant author’s previous research points to the lack choices their customers are making with of undisputable numerical threshold values, respect to the products and services retailers able to act as goodness of judgement’s are offering to them. This study aims to jackknives. This paper illustrates a address the complexity of involving method of meta-analysis of the paired information from a big data environment. relationship between the previously two The responses provided by the participants mentioned indices. Building on the previous in this respect are representative of rapidly author’s experiments which revealed the varying customer’s choices. Furthermore, existence of a linear association in between participants evaluated their suppliers based consistency and compatibility, the concept of on each relevant criterion. The responses accuracy-usually applied in the medical

44 decision making, is adapted and extended to third section where results of several point sensitivity of accuracy in order to filter experiments are presented and the out the outliers in this association. Results goodness of the link between the indicate that although a threshold of 0.1 for consistency and compatibility is casted and consistency index(ratio) might be arbitrary, sorted favorable, in terms of point sensitivity scale dependent and also dependent on of accuracy. Conclusions of this paper point various types of senses involved in a decision to a broad range of applications of this idea, making problem (vision, somatic sensation, once validated. logical /mathematical , existential, linguistic ) filtering judgments through the accuracy 2. Pairwise Comparison Matrices in test diagnostic as adapted hereafter offer a Accordance with Saaty's scale more reliable tool in heterogeneous decision *Michele Fedrizzi, University of Trento, making problems. The paper is organized as follows. In the first section is Italy, [email protected] provided a critical analysis of the literature James Fleming, University of Trento, regarding the appropriateness of the usual Italy, [email protected] 0.1 threshold for the consistency ratio as well Alexandra Caprila, University of Trento, as for the various types of compatibility Italy, [email protected] indices available to measure the closure of a derived priority vector to a certain “true This study investigates the number of “priority vector, which is determined outside possible consistent Pairwise Comparison the current process of decision making. In Matrices (PCMs) where all matrix elements this context, the scale dependency of the belong to Saaty’s scale (the set of integers consistency ratio, methods of improving on from 1 to 9 and their reciprocals). Many consistency with effects on the priority studies consider PCMs in Saaty’s scale, but vectors as illustrated in the relevant scientific the problem of determining the total number literature and brought together with author’s of these matrices that are consistent has own calculations about the chances to fulfill only been mentioned briefly by a single study consistency as the dimension of the decision (J. A. Alonso and M. T. Lamata, 2006). The matrices increases. About the compatibility number of n x n PCMs with elements in the index, aside of Saaty’s one, it is also scale is 17 raised to the power n(n-1)/2. reviewed Garruti compatibility index and However, of these matrices, only a small another slight adjustment of Saaty’s fraction satisfy the criteria for consistency. compatibility index is introduced. A Our starting point was the fact that the rank comparison in between the three of a consistent PCM is one, and all of its rows compatibility indexes is performed through a are proportional to each other. In addition, all Monte Carlo experiment and the existent diagonal elements are equal to one. critical comparisons are re-interpreted. In Therefore, the elements of the first row the second section, the concept of test completely determine all other matrix accuracy, as it appears in medical decision elements. By inspecting all possibilities for making is presented, together with an the first matrix row, we determine 1. a original analytical determination of the method to generate all the consistent PCMs accuracy’s point sensitivity. The second in Saaty’s scale and 2. a closed-form formula section is devoted to the analysis of the for computing their number as a function of paired relationship between consistency and n. For n = 3, there are 4,913 PCMs in Saaty’s compatibility in the context of calibration scale, among which only 85 are consistent. curve, allowing for the evaluation of the Similarly, for n = 4, there are 24,137,569 linearity and homoscedasticity. The concept PCMs of which only 353 are consistent. For n of accuracy , as the most important aspect = 5, there are 2,015,993,900,449 PCMs of the validation of the calibration curve in and only 1,381 are consistent. Our results the medical domain is extended to the differs from those reported by J. A. Alonso concept of point sensitivity of accuracy and and M. T. Lamata (2006). We validated our in this section is proved how several formula for n = 3, 4 and 5 by generating all concerns in the detecting of outliers applying the PCMs in Saaty’s scale and checking the accuracy are favorable flicked with this new consistency conditions for each. In each analytical derivation of the accuracy concept. case, the check confirmed the result of our This new concept is further applied in the formula. For n equal or greater to 6, the

45 number of PCMs to be checked becomes These indices allow assessing the degree of intractable. inconsistency of the incomplete pairwise comparisons matrix. Thus, they can be used References Alonso, J. A., & Lamata, M. T. to asses the credibility of such a ranking. (2006). Consistency in the analytic hierarchy Theoretical considerations will be process: A new approach. International accompanied by Montecarlo experiments Journal of Uncertainty, Fuzziness and during which we study the performance of Knowledge-Based Systems, 14(4), 445– the modified inconsistency indices. 459. Saaty, T. L. (1977). A scaling method for priorities in hierarchical structures. 4. A Set Cover Problem Approach for Journal of Mathematical Psychology, 15(3), Improving Consensus and 234–281. Consistency in Large (AHP) Group 3. Inconsistency of Incomplete Decision Making. Pairwise Comparisons Matrices *Jorge Romero, Universidad Jorge *Konrad Kułakowski, AGH UST, Poland, Tadeo Lozano, Colombia, [email protected] [email protected] Felix Antonio Cortes Aldana, The pairwise comparisons method is the well-known procedure of ranking creation. In Universidad Nacional de Colombia, the basic model, experts compare each of Colombia, [email protected] the two alternatives, thus, form the matrix of Monica Garcia-Melon, Universitat comparisons (the pairwise comparisons Politecnica de Valencia, Spain, matrix), next, they use the mathematical [email protected] methods to prepare the final ranking. Probably the most popular method based on The purpose of this work is to show a way to this approach is the Analytic Hierarchy improve consensus and improve consistency Process (AHP). Nevertheless, such a in group decision making problems, using procedure, although simple, has two AHP, matrix similarity comparison and disadvantages. The first of these is simulated annealing. The main goal of this inconsistency. As experts do the individual work is contributing to the challenges and assessments, they may not always be open questions proposed by Cabrerizo et al consistent. The level of inconsistency of in 2013 highlighting the following areas: paired comparisons is determined by so- counseling, visualization and verbalization of called inconsistency indices. It is widely the process, dynamic decision contexts, and accepted that the higher the inconsistency, persuasion. A hybrid approach is proposed to the lower the credibility of the ranking. The address the issue of dealing with a large second disadvantage of this approach is the group of decision makers or several need to compare all alternatives against stakeholders, and then search for the problem consistency changing the minimum each other. On the one hand, this may require much work, on the other hand, it may number of judgment values, also evaluate not always be possible. Therefore, more and the agreement of their judgments in every more often in practice, only a part of all pairwise comparison and for the whole possible comparisons are made, and the model. The evaluation of the agreement level ranking is calculated using incomplete is performed in order to rank the consensus pairwise comparison matrices. Despite this level or level of agreement of every decision popularity, it seems that the problem of the maker to all others, and calculate a global inconsistency of incomplete pairwise consensus index for the problem. Once the comparisons matrices is not well specific pairwise comparisons with no represented in the literature. The presented agreement are identified, a persuasion work aims to (at least to some extent) bridge attempt is performed in these comparisons, trying to change the minimum number of this gap. In the paper, we discuss several well-known inconsistency indices such as judgments in order to reach a stablished Saaty’s CI, GCI (Geometric Consistency consensus level. Decision model start using Index), GWI (Golden-Wang Index) and others, AHP to construct the decision matrices for all and present their extensions to the decision makers, given that this work uses incomplete pairwise comparisons matrices. AHP a consistency analysis is performed in every triad of values for all pairwise matrix, 46 based on the work proposed by Kendall and Juan Bravo, Universidad del Valle, Babington in 1940 “On the method of paired Colombia, comparisons” and extended by Kulakowski [email protected] in 2018. This initial approach for improve Pablo Manyoma, Universidad del Valle, consistency consists in a variation of set Colombia, covering problem proposed by karp in 1972 called triad covering problem, and [email protected] considerate every circular triad as an According to the International Federation of individual point of consistency (according to Red Cross and Red Crescent Societies, a the axioms given by Kendall in 1940 and disaster is a sudden and calamitous event kulakowski in 2018). The algorithm for that seriously disrupts the functioning of a perform evaluations in every pairwise community or society. This event causes comparison matrices are performed by using human, material, economic and the JULIA language. Once have consistency environmental losses, which exceeds the in the problem, the first consensus index is capacity of the community or society to use calculated and if don’t have reach the their own resources and therefore generates desired consensus index, a second algorithm an appeal to Humanitarian Assistance. In all is performed to search the minimum number the work of humanitarian assistance there is of changes in decision makers judgments in a very important factor called donation. A order to reach the appropriate level of donation is a voluntary delivery of one or consensus for the problem (the consensus more goods, by a donor, which may or may control is based on the previous work of not be accepted by the donee. Donors can be Romero-Gelvez and Garcia-Melon in 2016). different humanitarian organizations, The algorithm for consensus control is companies or individuals, who are performed also with JULIA programing encouraged to donate for two fundamental language. The validation of this approach is reasons: by request of collaboration or by proved in three different scenarios: The first natural reaction to the event. Donations have application is ranking the main airports of an appearance before, during, and after the Colombia dealing with 85 decision makers, disaster. In the "before" of disaster, standard the second one is an environmental kits are made with items such as kitchen application for ranking eco-tourism in utensils, tools, clothes, hygiene and others. Colombian Amazonia dealing with 8 decision Once the disaster has occurred, the entities makers. And the last one is an educational must give a quick response to the affected application to improve performance in population, complementing the aid inventory undergrad students of engineering with the donations that are starting to be programs, dealing with 30 decision makers. received at that moment. People have an unfortunate perception of what should be MON-3-C donated when a disaster occurs, since most Invited Session: Building MCDM/A believe that it generates more help by Models: Practical and Methodological sending everything it can as quickly as Issues possible. These types of perceptions can generate large amounts of waste both from Monday 13:50-15:30 - Room: Dolmabahçe donations and from the resources that must Palace be used to receive, maintain and transport inappropriate or unsolicited donations. To Chair: Caroline Mota, Cristiano Cavalcante mitigate this negative impact in some way, it 1. Quantification of Acceptability is important to be able to identify when a Level for Donations in Case of donation is efficient and can really help. Therefore, through the characterization of Natural Disasters appropriate donations and the *Adriana Arias, Universidad del Valle, establishment of fundamentals criteria, it is Colombia, intended to develop a methodology to [email protected] quantitatively qualify the level of acceptability of donations in a case of natural Daniela Morales, Universidad del Valle, disaster (flooding) in the city of Santiago de Colombia, Cali (Colombia). In Colombia, the National [email protected] Disaster Risk Management Unit has

47 established the different topics that make up multiple criteria. In majority of MCDA tasks, the assistance: water and sanitation, food the decisions are based on subjective aid, non-food aid (cooking kits, hygiene and information about a decision maker’s bedding), temporary accommodation and preferences. Numerous descriptive studies health days. For this research, only food and examined different ways of eliciting such non-food aid issues will be consider, since human preferences and their comparative they are the ones that are most closely consistency. As a rule, they conclude that related to donations and donors. preferential information from the decision Quantification of acceptability level of makers in the ordinal form is more consistent donations is made through five criteria: and thus more reliable than one obtained by Proximity to expiration, use time, quantity, using quantitative (interval or ratio) scales. size and product appearance. These criteria Verbal Decision Analysis (VDA) is a allow to evaluate all the basic elements that framework for multicriteria decision making make up a food kit, toilet kit, kitchen kit and methods, based primarily on using subjective bed kit. The AHP multicriteria methodology preferences in the verbal, ordinal form; and was developed to obtain the criteria this study reviews the main ideas and prioritization. This was achieved thanks to methods developed under this framework. the participation (surveys) of possible Oleg Larichev and Helen Moshkovich in affected people, donors and experts from 1997 first proposed and outlined VDA some entities. Then the donations receive a approach with first three methods – ZAPROS, score depending on the theoretical ORCLASS, and PARK based on VDA acceptability they have. Finally, the resulting paradigm. The main idea that a decision percentages give a global score, which is maker should be allowed to express his/her equivalent to the level of acceptability of the evaluations and preferences in a verbal form donation. All the above is done with the help and that final decisions should be based of a macro in Excel. Once the prioritization entirely or primarily on these verbal ordinal has been obtained, it is evident that when preferences without transformation into a the product is perishable, a large part of the quantitative form in some arbitrary way. judgment is concentrated in the proximity to Another important idea of VDA was its expiration. In the absence of this criterion, incorporation into the decision analysis the appearance of the product will take on a special dedicated procedures for verification greater weight. In addition, in all cases, it is of the consistency of a decision maker’s well known that quantity criterion acquires preferences. Following these principles great importance in donations evaluation. makes the decision process transparent to a The methodology is proposed as an decision maker by providing easy to follow educational tool that generates a culture of explanations of the final results. The first "good donation" in the city. When validating three methods, based on VDA approach, the methodology, there is a change in were developed for different types of tasks. people's perception of what a donation is and Method ZAPROS was designed for rank what is appropriate to donate, as well as ordering alternatives, ORCLASS was feeling more secure and more willing to help developed for an ordinal classification of others. alternatives, while method PARK helps to select the best alternative. As VDA approach 2. Comparative Analysis of VDA becomes more popular, many new methods, Methods and Techniques based on VDA principles, were proposed as practical tools to solve multicriteria decision *Alexander Mechitov, University of making problems, including ARACE, CLARA, Montevallo, United States, etc. In this study we outline the main [email protected] peculiarities of these methods, classify them, Elena Moshkovich, University of and discuss the current trends in VDA Montevallo, United States, development. [email protected]

Multiple Criteria Decision Analysis (MCDA) deals with decision making tasks which require to find the best or a group of the best alternatives while taking into consideration

48

3. A Multicriteria Model to Evaluate approach to evaluation of maintenance Opportunistic Maintenance Policies policies. Subject to Imperfect Inspection 4. A Spatial Multiple Criteria Mônica Marsaro, UEMA, Brazil, Decision Model for Vulnerability [email protected] Analysis Alexandre Alberti, UFPE, Brazil, Ciro Figueiredo, Universidade Federal [email protected] de Pernambuco, Brazil, Crıstıano Alexandre Cavalcante, UFPE, [email protected] Brazil, [email protected] *Caroline Mota, Universidade Federal This paper presents a multicriteria approach de Pernambuco, Brazil, to evaluating opportunistic maintenance [email protected] policies subject to imperfect inspection. This study highlights the benefits of using a Opportunistic maintenance refers to the spatial MCDA approach to evaluate the strategy of carrying out preventive vulnerability of places at crime level. The maintenance actions on a particular sub- model was supported by a Dominance-based system at the time of an opportunity rough set approach to join preference generated by a failure or preventive action on learning and Geographic Information another sub-system that is part of the same Systems. The approach requires a priori system, which can lead to savings in terms of knowledge from decision-makers for holistic cost and downtime. In this work we present assessment to analyze individual results. a delay-time model for a hybrid inspection Additionally, we explored an approach for and preventive maintenance policy subject aggregation of those individual results for to imperfect inspections, and which also obtaining a final output recommendation. considers the possibility of opportunistic The approach aims to decrease the cognitive maintenance after a certain time of effort of decision makers, in which their operation. Since maintenance decisions preferences is assessed in an interactive and often involve objectives other than cost incremental way. We present an application minimization, given the possibility of impacts of the proposed approach to evaluate places on non-financial dimensions, multicriteria according to its vulnerability to crime decision-making approaches can be occurrence and to identify areas that appropriate for these contexts. In this paper deserve more resources to combat crime. As we explore the impact of imperfect a result, the model may help policy-makers inspections on the maintenance policy and planners to develop public policy performance for two criteria: expected cost per unit of time in the long term (cost) and interventions. mean time between failures (MTBOF - a good MON-3-D measure of reliability) resulting from the adoption of the maintenance policy. These Tutorial: Improve Your Decisions by criteria are conflicting because there is no Learning and Experiencing the ANP maintenance policy that presents the Best Practices- Part I optimum performance for both, thus a multicriteria approach based on the Multi- Monday 13:50-15:30 - Room: Topkapı Attribute Value Theory (MAVT) was Palace considered for the evaluation of Chair: Birsen Karpak maintenance policies. MAVT was chosen because compensatory rationality is Improve Your Decisions by Learning compatible with the problem, and even when and Experiencing the ANP Best there is no complete knowledge of the state Practices- Part 1 of 2 of nature of the problem, the method can bring important insights. A numerical *Orrin Cooper, University of Memphis, application is presented, and the results United States, [email protected] show the impact of the imperfect inspection, This tutorial is an Analytic Network Process the mistakes that can be made when this (ANP) Best Practices experiential learning factor is not considered, as well as the activity because we learn best by doing. importance and adequacy of the multicriteria 49

Beginners with no prior experience will make 1. Robustness Indicators for Multi- their first ANP decision. Experts, and those in Objective Integer Linear between the two extremes, will acquire valuable tools and techniques to improve the Programming validity of their ANP decisions. Attendees *Michael Stiglmayr, University of should come prepared with an interesting Wuppertal, Germany, decision, big or small, that has at least 3 [email protected] alternatives and 2 clusters, i.e. groupings, of José Rui Figueira, Technical University criteria, with multiple criteria in each cluster to build a decision model during the tutorial. of Lisbon, Portugal, The fundamentals of the ANP will be [email protected] addressed, just in time, as we build our Kathrin Klamroth, University of individual decision model. The ANP Best Wuppertal, Germany, Practices will serve as our checklist to make [email protected] sure that we address the critical components Luís Paquete, University of Coimbra, of an ANP decision. A review of ANP studies, Portugal, [email protected] that were published in the Social Science Britta Schulze, University of Wuppertal, Citation Index (SSCI) over a one-year period, Germany, [email protected] revealed that over half of the published ANP wuppertal.de models had errors or omissions that were serious enough to bring the validity of the Decision uncertainty is a robustness concept published model into question. These in continuous single and multi-objective findings contributed to the development of optimization, which takes into account the the ANP Best Practices. The most common or inaccuracy of the implementation of a critical omissions will be covered in greater solution (also called implementation error). detail at the specific times that they should Due to technical limitations a solutions might be addressed when we are building our not be feasible for implementation. decision models. Tools and techniques will Implementing a solution in the neighborhood be provided to help decision makers: to of the computed optimal solution can lead to check for the convergence of the a smaller or larger loss in solution quality. To Supermatrix, to identify and adapt a disjoint limit this loss one can, for example, optimize Supermatrix, to recognize and accurately the worst solution in the neighborhood capture the desired level of dependency in (minmax). Eichfelder et al. (2017) investigate the Supermatrix, to test and improve the decision uncertainty for continuous Coherency of the Supermatrix, and to multiobjective problems using set-valued conduct and report meaningful sensitivity optimization problems as robust analysis. By the end of this tutorial, beginners counterpart. We transfer this concept to will have “experienced” an ANP model from multi-objective combinatorial optimization beginning to end; and experts will have more problems. In the discrete context decision tools and techniques to improve the validity uncertainty is due to the fact that certain of their decisions. items (edge, element, ...) of a Pareto optimal solution might turn out to be unavailable in Please bring your own laptop to this special the implementation phase. Instead of session reoptimizing the problem one is interested in MON-3-E repairing a solution by substituting the flawed item. Thus, the originally selected Contributed Session: Recent Advances Pareto optimal solution is substituted by a in Multiobjective Optimization neighboring solution. The neighborhood structure is thereby defined by a Monday 13:50-15:30 - Room: Hagia Sophia combinatorial adjacency structure of the Chair: Petra Weidner problem. We propose robustness indicators, which are based on those neighborhoods. They are mainly of two types, cardinality indicators and quality indicators. In a first case study we evaluate the robustness indicators on randomly generated instances of the cardinality constrained knapsack

50 problems. Furthermore, by using these Moreover, when the decision maker has an indicators as a quality measure for the aspiration vector, then we can use reference representation problem we construct point methods where the objective is to find representative subsets of the non-dominated the solution whose image minimizes the set that are robust against decision distance to the aspiration vector. For set- uncertainty. based robustness, that is finding the solution whose set of worst cases is the ‘closest’ to 2. A New Algorithm for Optimization the aspiration vector. The proposed method over the Efficient Set has three key components. First, since the image of the solutions is a set, we use the *Kahina Ghazli, USTHB, Algeria, averaged Hausdorff distance to measure the [email protected] distance from the image of a given solution Nicolas Gillis, Université de Mons, to the aspiration vector. This formulation Belgium, [email protected] reduces the problem to a single-objective Mustapha Moulai, USTHB, Algeria, optimization problem. The averaged [email protected] Hausdorff distance has been used by the evolutionary multi-objective community to Optimization over the efficient set of a multi- compare the solutions found by different objective optimization problem is among the algorithms, and it has also been used as a difficult problems in global optimization selection mechanism for evolutionary because of its nonconvexity, even in the algorithms. Next, for every solution in linear case. In this paper, we propose a decision space, it is required to find the set numerical method to tackle this problem of worst-cases. For this purpose, we use the when the objective functions and the weighted Tchebycheff method to find a good feasible set of the multi-objective representation of the set. Finally, to find the optimization problem are convex. This solution that minimizes the averaged algorithm penalizes progressively iterates Hausdorff distance, we use a global search that are not efficient and uses a sequence of method that enables to avoid local optima. convex nonlinear subproblems that can be Namely, we use the differential evolution solved efficiently. The proposed algorithm is algorithm (DE/rand/1/bin) as the search shown to perform well on a set of standard engine. DE has been extensively used when problems from the literature, as it allows to tackling single-optimization problems in obtain optimal solutions in all cases. continuous spaces with good results. We test 3. A Differential Evolution Reference the method on an academic test function Point Method for Set-based based on the bi-objective Lamé super- spheres with linear, convex and concave Robustness by Means of the fronts. The results show that the method is Averaged Hausdorff Distance capable of finding good solutions from the problem at hand given different aspiration *Carlos Hernandez Castellanos, vectors. University of Oxford, United Kingdom, [email protected] 4. Scalarization Depending on the Sina Ober-Blöbaum, University of Purpose of Multiple Objectives in Oxford, United Kingdom, sina.ober- Optimization [email protected] *Petra Weidner, HAWK HHG, Germany, In this work, we propose a method to find [email protected] efficient solutions in the context of set-based robust multi-objective optimization. In this When looking for the best solution in setting, a solution in decision space maps to practical applications, multiple objectives a set, which represents the worst possible can come into existence by different reasons. outcomes given the set of uncertainties. They can express values of attributes, Thus, the task is to find the set of best-worst various scenarios in decision making under solutions. This set is highly attractive when uncertainty or evaluations of multiple the decision maker has an aversion towards decision makers. This should be taken into risk and would like to perform the decision consideration for the choice of a scalarizing making from a worst-case approach. problem. This aspect is discussed in the presentation. Moreover, it is shown that the 51 different scalarizations can be visualized in company-driven measures to develop spatial the same framework, which also delivers structures. As a result, there is no statements about their properties. The appropriate decision-making approach of presented framework is an extension of the location planning in metropolitan regions, scalarizing method by Pascoletti and that fulfills the requirements with respect to Serafini. It is applied to the Hurwicz rule for regional dynamics. There are three types of decision making under uncertainty, to relevant regional dynamics. First, dynamics problems with different weighting vectors in regional location planning includes over originating from multiple decision makers the time changing corporate location and to decision problems with domination requirements. These generally increase due sets which are not necessarily cones. to organizational, economic and technological advances. Second, the MON-3-F characteristics of regional location factors Contributed Session: MCDM for Facility are not temporally constant but underlie Location and Logistics changes over time. On the one hand, these changes result from municipal decisions in Monday 13:50-15:30 - Room: Basilica the course of location developments, Cistern influencing the characteristics of regional location factors. On the other hand, Chair: Özay Özaydın companies are able to allocate appropriate 1. Multi-Criteria Facility Location development measures to improve the Planning and Development under characteristics of regional location factors in a target-oriented manner and according to Consideration of Municipal their own location requirements. Third, Developments in Metropolitan municipal and company induced location Regions developments may cause synergetic or conflicting interdependencies between *David Kik, TU Braunschweig, Germany, regional location factors. By now, these [email protected] dynamics are not considered in regional Matthias G. Wichmann, TU location planning. Thus, an appropriate Braunschweig, Germany, model formulation for regional location [email protected] planning considering development Thomas S. Spengler, TU Braunschweig, measures and regional dynamics is missing. Germany, t.spengler@tu- In this contribution, a multi-period weighted braunschweig.de goal programming model for the integrated location planning and development in The increasing development of metropolitan metropolitan regions is developed. The regions and the associated changes of integrated decision model allows to spatial structures are the main simultaneously select a compromise-optimal consequences of globalization and location and to generate a strategic urbanization trends. Metropolitan regions development plan for the allocation of are characterized by a multitude of different optimal company-driven measure sets to spatial structure types. These differ develop locations in a target-oriented significantly regarding the quality and the manner. Furthermore, the model takes into development of economically relevant account that location construction regional location factors. However, the investments as well as investment payments differences of spatial structures offer for location development do not exceed a potentials to increase a company’s restrictive overall budget. Comparing existing competitiveness and to ensure a sustainable location planning models with this novel company’s success. In order to exploit the integrated location planning model, findings identified potentials, companies must of the latter show a better achievement of consider relevant future dynamics in the the decision-makers targets by course of location developments. To date, approximately 43% due to the company- the dynamics in regional location driven measure allocation. Therefore, developments are not yet considered in significantly better location factor existing facility location planning characteristics are achieved according to the approaches. This especially holds for company’s requirements and long-term

52 optimal location decisions under ambulances, the police and other entities consideration of municipal developments in involved in emergency situations would face metropolitan regions can be ensured. difficulties in reaching the most affected areas and in acting effectively to limit the 2. Using Multicriteria Decision losses in post-disaster phase. For these Making in Emergency Situations – reasons, the preparedness is an extremely Case Study: Bucharest City important component for risk reduction. The present paper aims to apply a multi-attribute *Diana Popovici, University of decision making for identifying suitable Bucharest, Romania, areas in the city, that can be used as local [email protected] centres for conducting the rescue missions, Iuliana Armas, University of Bucharest, for sheltering the population and for giving Romania, [email protected] the first aid in the post-disaster phase. The Alexandru Gavris, The Bucharest criteria considered are those regarding the University of Economic Studies, physical vulnerability of the city, the social Romania, [email protected] vulnerability and the capacity of intervention of the authorities. The results can be used by Dragos Toma-Danila, National Institute the local authorities for improving their for Earth Physics, Romania, preparedness and for limiting the human [email protected] losses. Bucharest City is the capital of Romania and 3. Prioritization of Areas for High it’s one of the most vulnerable European capitals to earthquakes. The vulnerability Volume Loading and Unloading. towards earthquakes resides from its Case: Santiago de Cali geographical position, the age of the *Eduar Aguirre, Universidad del Valle, buildings and the high population density. The earthquakes that affect Bucharest City Colombia, have the origin in Vrancea Region. These [email protected] earthquakes have the hypocentre at depths The freight consolidation´s is defined as the between 70 and 200 km and can reach a process of grouping different shipments of magnitude of M7.0 – M8.0. The waves suppliers into a large shipment within a propagate on a northeast – southwest consolidation center. The main objective of direction from the epicentre. The major freight consolidation´s is to reduce the total earthquakes that affected Bucharest in the cost of transport between an origin and a last century were registered in 1940, 1977, destination. Some authors classifies in three 1986 and 1990 causing many victims and forms of consolidation: consolidation of economic loses. Based on the scientific inventories, consolidation of vehicles and evidences, it is very probably that a new consolidation of terminals. Another authors major earthquake is yet to strike the city. make a similar classification, but adds a form Many of the city’s buildings are very old, with of consolidation called independent, in which damaged structure, that passed through small shipments are attended directly to many major seismic events, being a real each client independently. To this danger for their inhabitants and for their classification denominates consolidation of surroundings. The city is also highly shipments, route of vehicles and populated, many neighbourhoods developed consolidation of network respectively. in the communist period have blocks of flats Urban freight transport is an extremely with 4, 8 and 10 storeys. During the working important and quite disturbing activity. hours, to the residential population is added Increasingly, one observes efforts to the population that lives outside Bucharest, measure and control the movements of but that works or studies in the city. Another cargo within city centers. The vast majority of factor that contributes to the high public sector initiatives aimed at urban cargo vulnerability of the city are the narrow streets aim to reduce their negative social and that surround the highly dense areas. All of environmental impacts, which are typically these added to the more and more intense the result of freight vehicular activity. This traffic, can contribute to the scene of a approach is a natural response to the disaster in case of a new major earthquake participation of cargo vehicles in the occurrence. The firefighters, the 53 generation of negative externalities. consideration. Initially, a preliminary survey Meanwhile, urban logistics is a fundamental is conducted using Web of Science, Scopus, aspect for the development of a social and Science Direct databases for 2007- environment and more when you have a high 2019. Keywords “logistics,” “assessment,” volume of commercial activity and freight “performance,” “evaluation,” “supply chain,” transport, for a broad population density; so “transportation” are used as corresponding urban logistics is defined as the process search strings. The resulting articles are through which the private transportation and further re-classified into 15 groups, namely: the logistics activities in urban areas are appraisal, big data, cost-benefit analysis, optimized; considering traffic, congestion collaboration, decision, infrastructure, and energy consumption within the structure optimization, performance, supply chain of a market economy. The Valle del Cauca´s management, service, simulation, region its located at southwest of Colombia, sustainability, transportation, competitive yours socio-economic development of the advantage, and scenario. Within this has been mainly determined by coffee, sugar classification, social network analysis is and, agricultural production of inputs for performed among the mentioned articles agro-industry.. Additionally, the city of Cali using ORA software. Degree centrality is used has different areas with commercial, to find the articles having the most residential, academic and health activities, significant number of direct links to-and-from some clustered and others mixed together. other articles. This is used as an indicator for This condition makes the cargo vehicles leading articles. Betweenness centrality is circulate throughout the city looking to used to find the amount of control that paper deliver their goods as close as possible to the has over the others. This shows whether final distribution, generating congestions there is an article that controls access to all and accelerated deterioration of the internal other articles in the network. Eigenvector roads due to heavy traffic. Considering the centrality is used to measure how well a territorial analysis and the concept of urban paper is connected to other well-connected logistics, with the help of the Multicriteria papers. Additionally, the density of the Decision Analysis (MCDA), such as Analytic network, its diameter, its degree Hierarchy Process (AHP) and Technique for centralization are analyzed. Degree centrality Order Preference by Similarity to Ideal is used to specify the relative dominance of Solution (TOPSIS) methodology, it is intended a single article over all other articles in the to provide a ranking of proposals on areas in network. Networks with a high degree the city with current and future possibilities centralization are subject to fragmentation if to be considered as a high volume loading these highly connected nodes are removed. and unloading areas, considering differents The density represents the ratio of the criteria, include in Economic, Technical and number of links presents the total number of Social dimensions links possible: the higher the density, the more complex the structure with many links. 4. Determining Leading Research on The diameter measures the connectedness Transportation and Logistics of the network, and it represents the amount Performance through Social Network of effort needed for information to move from Analysis and TOPSIS one end of the network to the other. Through this analysis, the changing trends in the *Özay Özaydın, Doğuş University, performance assessment methods are Turkey, [email protected] revealed, and finally, this paper proposes Füsun Ülengin, Sabancı University, new research areas in this field. Finally, Turkey, [email protected] TOPSIS analysis is performed to indicate the most important articles and therefore to The primary aim of this research is to underline the recent research areas. The determine pioneering literature on logistics criteria used in TOPSIS is composed of performance appraisal and evaluation. centrality measures as well as the impact Although the most common assessment factor of the corresponding journal. system is the cost-benefit analysis, it lacks to take temporal dynamics, qualitative logistics performance dimensions such as speed, on- time performance and quality into

54

MON-4-A 2. Multiple Criteria Benchmarking: A Contributed Session: Advances in Strategy of Improvement MCDM Theory *Jean-Philippe Hubinont, Université Monday 16:00-17:40 - Room: Galata Tower Libre de Bruxelles, Belgium, [email protected] Chair: Salvatore Greco Yves De Smet, Université libre de 1. Integrating MCDA with Strategic Bruxelles, Belgium, Planning [email protected] José Rui Figueira, Technical University *Theodor Stewart, University of Cape of Lisbon, Portugal, Town, South Africa, [email protected] [email protected] This paper is related to the benchmarking We have previously made the assertion that problematic in a multiple criteria context MCDA deals with aid or support for decisions where the alternatives of a given set are that are complex enough to “matter”. Many compared on the basis of a given set of reported examples of MCDA applications, criteria, through any method, which allows to however, are decidedly operational in nature provide a partial or complete ranking when and some border on the trivial. The thrust of applied to a performance table. In this this paper is to urge the MCDA community to context, a Decision Maker (DM) is willing to become more involved in messy strategic improve the performances of a certain level problems. Of course, many such have alternative. The aim of this study is to provide been reported, but even then very often the her/him with an improvement strategy that MCDA content has focused more on the final consists of: A final performance for each selection phase, of choice between action criterion that should and could be reached, alternatives, with very little information on and step by step guidelines to reach the final the origin and development of the criteria performances. Moreover, our method alternatives. We have previously argued that is built with the help of a theoretical MCDA can and should be viewed not only as framework that we develop that aims at an analytical tool, but also as a problem consolidating and defining notions that are structuring method (PSM). As such, the used in the benchmarking literature but not approaches and tools of MCDA have clearly settled or defined, in our opinion. relevance as a guiding principle throughout the entire process of strategic planning. In 3. Experimental Evaluation of order to clarify this assertion, and to provide Multiple Criteria Utility Models with a framework for the implementation of MCDA Veto Related Preference Structures in this context, we follow the three-phase structure of the strategic planning process as *Andrej Bregar, Informatika d.d., identified by Mintzberg, namely: (a) Slovenia, [email protected] Identification (recognising problems and A fundamental and widely applied approach diagnosing cause effect relationships to decision-making is the multi-attribute leading to action opportunities); (b) utility theory. Because it aggregates Development (generating or designing preferences in the compensatory manner, it potential solutions to the problem or crisis); has been recently extended with the concept (c) Selection (comparison, screening and of veto function, which has been adopted choice of actions). We examine the roles and from the outranking approach and models types of MCDA intervention that apply at full or partial non-compensation of each phase, and discuss the naturally unsatisfactory preferences. Within the scope dynamic and iterative nature of the process. of our past research work, the underlying MCDA can and should play a guiding and methodological foundations and approaches integrative role throughout. The presentation to express non-compensation in the utility concludes with an illustrative example, which based multi-criteria decision models have is a hypothetical case but broadly based on a been systematically studied and analysed, real-world situation in water resources with the focus on the veto criterion, veto planning. function and aggregation operators. The

55 main contribution of the presented research unable to discriminate alternatives as work pertains to the investigation and efficiently as either ROC or RS weights. The evaluation of forms and properties of veto obtained results indicate that the functions in multi-criteria decision models specification of veto can enhance the that incorporate utilities as well as efficiency and credibility of decisions. discordance related information. Particularly, Several quality factors potentially improve, the risk aversion of veto functions is studied including the accuracy and validity of results, and correlated with the risk aversion of utility ability to discriminate optimal from functions, which are aggregated in the same suboptimal alternatives, and robustness of multi-attribute model and exhibit common judgements. This is a consequence of the complementary preference structures of the fact that additional preferential information decision-maker. Outcomes of risk averse, increases the expressiveness and risk seeking and risk neutral veto functions completeness of quantitative models. are analysed and compared. The study aims Finally, the effect of veto functions in multi- to assess the influence of risk aversion on criteria utility models is analysed in the decision, to identify possible anomalies comparison to the notion of constraints in in preference structures, to determine the multi-objective optimization models. A suitability of different risk aversion formats couple of applications are assessed. and intensities for various problem settings, and to derive key characteristics. The 4. ELECTRE Score: A first Outranking experimental model is based on the Based Scoring Method simulation study. Several evaluation factors *Salvatore Greco, University of Catania, are observed, such as the ability to efficiently discriminate alternatives, richness of output Italy, [email protected] data, extremeness of results, validity of José Rui Figueira, Technical University results, correctness and relevance of of Lisbon, Portugal, judgements, and robustness. Partial results [email protected] of the study have already been presented in Bernard Roy, LAMSADE - Paris- the past. However, the scope has been Dauphine University, France, limited to ranking, additive aggregation and [email protected] complete rank-orders only. In this presentation, the completed research is We present an outranking method to assign addressed. Different problem solving a score to a set of actions. It is a method of problematics, aggregation models and types the Electre family, and we will call it of rank-orders are considered. The ELECTRE-Score. Differently from the Multi- simulation model is extended to deal with Attribute Value Theory (MAVT) methods, (1.) ranking and sorting, (2.) additive and ELECTRE-Score does not construct a value multiplicative aggregation operators, and (3.) function for each criterion, and then proceed complete, weak and partial rank-orders. to the aggregation into a single value. It, Based on the steepness, various shapes of rather, makes use of an outranking relation utility and veto functions are modelled that to make a comparison with reference sets of exhibit different risk aversion characteristics, actions, to which we assign a score with a ranging from very risk seeking and slightly deck of card method, and proposes a score risk seeking to neutral, slightly risk averse range for each action, instead of a single and very risk averse. Several specific value. Given the fragility of a single punctual scenarios are also defined to cope with score, the interval score assigned by the mixed, uniform, conflicting and proposed method seems a more robust way predominantly good/weak alternatives. The of proceeding. The sets of references actions outcomes of risk averse, risk seeking and satisfy the same properties of ELECTRE Tri- risk neutral veto functions are compared to nB. The fact of being able to use outranking standard ordinal preference specification relations makes it also possible to take into methods, such as ROC (Rank Order Centroid) account the imperfect knowledge of data and RS (Rank Sum) weights. It is shown that and to avoid systematic compensatory very risk averse veto produces similar results effects. Some fundamental theoretical as RS weights, while other forms of veto results guaranteeing the consistency of the approach ROC weights. On the other hand, method and an illustrative example are fully compensatory (utility only) models are provided.

56

MON-4-B this case study consists of the following steps: a brief literature review of recent Special Session: AHP/ANP Applications applications of MCDM methods; in Production and Manufacturing – identification of the motivation for this work sponsored by BORCELIK and decision of using AHP and ANP as two relevant approaches; selection of the Monday 16:00-17:40 - Room: Maiden's company and its current key decision Tower process; conception of the AHP model, Chair: Irem Duzdar Argun criteria and attributes; interviews with involved decision makers; processing of the 1. A Comparative Application of AHP collected data and discussion of the results and ANP in the Supplier’s Choice of outputted from the model. The authors a Capital Spare Part in a Small conducted interviews with the 3 senior Industry managers (maintenance, production and quality) connected to the aforementioned SP *Márcio Rodrigues, Technical University and its purchasing process. Furthermore, of Liberec, Czech Republic, historical data of a few Key Performance [email protected] Indicators (KPI) directly connected to this Eva Šírová, Technical University of process are incorporated into the model and Liberec, Czech Republic, their impacts are analyzed and discussed. Concerning the attribution of the weights, [email protected] Saaty’s fundamental scale was considered, The current reality of Industry 4.0 and but only using the odd scores for relative digitization increasingly requires a much importance (1, 3, 5, 7 and 9) and the reasons more accelerated and adaptive evolution in for assigning them by each decision maker business processes and enterprises. As a are also presented and discussed. The direct consequence, their key leaders are results obtained through the decision pushed to make decisions in a shorter time process using the AHP and ANP are frame and at a higher level of accuracy. This compared, since the interviewees also is only possible with the help of multi-criteria assigned relative weights between the 2 decision-making methodologies (MCDM), criteria. Some limitations of this research are already consolidated in the literature and due to confidentially required by the tested in companies of many economy company and there was also a lack of KPI segments globally, such as Analytic historical data from the purchasing Hierarchy Process (AHP) and Analytic department. Future research coming from Network Process (ANP). This article presents this article is possible, especially concerning a case study of the comparative application multi-criteria group decision making of these two approaches in a small industry (MCGDM) and hybrid MCDM models, since (production company dealing with these methodologies have many variations development and production of system and applications and could be combined with parts) for the decision between 3 suppliers of adaptive computational tools or with a capital spare part (SP) with 3 decision Machine Learning techniques, for example. makers inherent in the process. For this Those approaches could generate even more model, 6 attributes (supplier lead time, significant results and promote this decision- availability, spare part unitary purchase making process even more robust. price, part quality, aftersales customer service and sustainability) were grouped into 2. Methodology for Assessing the 2 hybrid criteria (Logistics-Economic and Degree of Automatism of an Quality-Environment). The main goal of this Operational Control Center in a research is to expand the knowledge base of Sanitary Company MCDM comparing outputs from 2 reliable methods, which there is still a lack in the *Claudio Macuada, Universidad de literature and can serve as a basis to a Santiago de Chile, Chile, further study and conception of hybrid [email protected] models combining multiple methodologies, Astrıd Oddershede, USACH, Chile, suitable to many other decision processes. [email protected] The research methodology used to conduct

57

Luis Quezada, Universidad de Santiago Khalid Akhtar, National University of de Chile, Chile, [email protected] Science and Technology, Pakistan, Pedro Palominos, Universidad de [email protected] Santiago de Chile, Chile, Mirza Jahanzaib, University of [email protected] Engineering and Technology, Pakistan, Companies indistinctly of their magnitude [email protected] and purpose, face a series of difficulties that The Global Facility Location (GFL) is a can affect the achievement of their fundamental but complex decision for top objectives. The different areas or activities, administration of the worldwide business as well as strategic initiatives, operations, firms, exploring the countries or nations as processes and projects may have individual potential future markets. The or strategic objectives, and these influence multidimensional nature of the GFL problem their results. This can have a direct or requires an extensive investigation of the indirect impact: The stakeholders, the criteria affecting the decision. There are only environment and society in general. In order a limited number of GFL studies available in to control the difficulties or problems when the literature which concurrently emphasize meeting an objective, it is necessary for on real world uncertainties, future horizons companies to establish an adequate risk and interactions of the stakeholders in GFL management process to allow them to issues. This work is a part of an on-going anticipate, manage and control the possible research project in which the agent based negative impacts and their consequences. modeling is merged with the analytical Hence, it is very important to ensure survival, hierarchy Process (AHP) to make a more prestige and financial stability. This study flexible and dynamic analysis of the GFL introduces the development of a problem by taking into account the real world methodology that uses a multicriteria uncertainties in a stochastic way. It focusses approach to select the procedure for on proposing an Agent Sensing Cycle (ASC) identification, analysis and evaluation of where the investor agents move to patches risks and thus determine the scope of the representing the countries and make an AHP automation of an Operational Control Center based decision to land on one of them to (CCO). In this case, the Analytic Hierarchy install a new production plant. These patches Process (AHP), is used, resulting helpful to are initially assigned the Country Objective find the risk management procedure that Functions (COF’s) which were calculated best fits the needs and for the problem from a Fuzzy-AHP study. A roulette wheel situation understanding. The expected aiming approach is proposed in the cycle of results will indicate whether it is cost- agent movements. The selection decision is effective to automate the CCO, in order to thus made based on the analysis of a reduce the risk of affecting the continuity of broader spectrum of influencing parameters supply of drinking water. The contribution is combined with the stochastic themes and a estimated in two aspects, introduces the focus on the uncertainties inherently present topic of risk management and provides a in information databases and expert useful methodology based on AHP. In this sentiments. Priorities of the investor agents context, it allows to extend this type of are changing during the simulation which analysis and evaluations to any area that make it a dynamic as well as stochastic requires it. Keywords: AHP, Risk MCDM analysis. The proposed study, Management, Sanitary Company. therefore, makes GFL results more reliable, comprehensive and exciting as compared to 3. An MCDM based Dynamic Patch other available traditional approaches. Sensing Cycle for Agent-Based Modeling and Simulation of Global 4. Prioritization of Digitalization Facility Location Criteria for the Product Design and Development *Hafiz Khurram Ali, Unviversity of Engineering and Technology, Pakistan, *Irem Duzdar Argun, Duzce University, [email protected] Turkey, [email protected]

58

Digitalization is defined as the integration of 1. Analysis of the Volumes of Foreign digital sources in all areas for improvement Trade for Turkey and the EU of digitalized firms and increasing their values. Digitalization is the development of Countries via Multi-Criteria Decision new job models, integrating the sources and Making Methods materials with new combinations to form *Ezgi Demir, Piri Reis University, Turkey, customer experiments, produce new goods, [email protected] and apply the technology to sources for using them efficiently. Integration of the European Union countries are the main digitalization with the organizational strategy trading partners of Turkey. Foreign trade is is very significant for the future performed in two ways as import and export competitiveness and success. The firms in terms of delivery of trading transactions. should have a strong digital strategy are Export has an important role for the known as the established the right way for development of national economies. Hence, their digitalization studies, and positioned raising exports and as well as reducing better at their digitalization route. In the imports are the vital targets for countries. literature survey, it is required that all of the The decisions and precautions made by goods developed and designed must be countries to achieve these goals form the adapted to the digital age. The digitalization basis of foreign trade policy. On the other success will be reached when the design is hand, international conjuncture affects focused on the customer experience. In this economic data. In this study, a sorting has empirical study, a literature survey is been performed among EU Countries and performed to categorize the criteria for Turkey by defining six main criteria which success in digitalization to define the factors affect the volume of foreign trade. These affecting the contributions of digitalization to criteria include both maximization and the product design. ANP is a practical MCDM minimization values. For the study, weights method and offers the advantages of of the criteria, has been determined via decision-making models, based on tangible Entropy method by using the 2018 data of and intangible factors. It allows creating 29 countries. Entropy method is preferred weights for different criteria to make the because of an objective method which does model stronger.In this study, the suggested not require an expert opinion according to model balances the economical, the data related to alternatives. After technological, infrastructural criteria determining the weight values of the criteria, considering product design and the countries were sorted by using the development. The aim is to prioritize these Copras method, which is the performance- criteria are evaluated by expert firms having indexed evaluation method. The Copras R&D departments. These results are method was used because it was the method analyzed using the ANP to prioritize the that provides the complex proportional digitalization criteria. By this way, it is assessment that incorporates the expected to establish a route of action for the maximization and minimization criteria. It is firms planning the digital applications to their expected that the study is going to present a products. new approach to the literature, as there exists no related study with the foreign trade MON-4-C data by using the Entropy based Copras Contributed Session: Multiple Criteria method. Decision Aiding 2. Multi-Criteria Assessment of Monday 16:00-17:40 - Room: Dolmabahçe Organizational Servitization Palace Readiness Chair: Isabelle D. Martins *Kerim Uygur Kizil, Istanbul Technical University, Turkey, [email protected] Esra Atac, Istanbul Technical University, Turkey, [email protected] Servitization, which is a concept that especially applies to the competitive

59 environment of developed economies, such kind of business model change. implies a shift in a firm’s business model Therefore, this study is especially aimed at from manufacturing to services. The the topis of organizational change to help the rationale behind such a shift can be divided manufacturers better understand their into two perspectives. First, firms may internal status-quo and make better choose to undergo such a change and decisions about their servitization strategy. It include services in their product-based is certain that multiple criteria need to be portfolios because they would gain financial assessed before making the final decision benefits. This is mainly because the non- about an organization’s readiness for transactional based nature of services can servitization. In this study, to determine the increase turnovers significantly and stabilize related criteria, several organizational the revenues. The second perspective is diagnosis models are examined including regarding the strategic and marketing McKinsey’s 7S model, Galbraith’s Star Model benefits. Competitive strategies that are and Burke-Litwin Model. All of these models purely based on products -such as enhancing consist of different elements that can be and innovating the products- do not lead to a used to explain the current situation of an long-term advantage nowadays since such organization. While assessing the models, efforts are becoming more imitable due to the ones with an open system approach were fast technological advancements and prioritized. The Burke-Litwin Model was decreasing product life cycles. Offering prominent in that matter. Based on this services along with the physical products model, different criteria categories are enables a long-lasting competitive selected which are environment, culture, advantage for the firms. Therefore, leadership, current business model, and so manufacturing firms that adapt to services on. A tool is also designed based on a are more likely to gain benefits through modification of the weighted sum model. By increased differentiation of offerings and using this tool, it is possible to ask questions customer satisfaction. On the other hand, to an organization member to capture the the process of servitization has its own big perceived benefits of the criteria. As a final challenges. A significant question here is outcome, the tool presents a decision of again financial. For example, the concept of whether the particular organization is ready service paradox implies that a firm wishing to for the servitization process or not. Unlike expand its service business confronts higher several similar tools in the literature, this costs for the benefit of more diverse service designed tool utilizes from multi-criteria offerings, yet it fails to get the expected decision making approaches and analyzes financial returns. Another challenge of the servitization readiness from an servitization is related to the differences organizational point-of-view. between goods and services. Manufacturing firms are often inexperienced in managing 3. A Review of the Multicriteria the significant characteristics of services, Decision Analysis Applied to Oil and which fundamentally are intangibility, Gas Decommissioning Problems variability, inseparability and perishability. Some other critical challenges include the *Fernanda Moraes, SAGE/COPPE, organizational ones, for instance, the Brazil, [email protected] transformation of the organizational culture Isabelle Martins, SAGE/COPPE, Brazil, and the organizational structure to fit better [email protected] the new business model. Considering both Giselle Tavora, LTS/COPPE, Brazil, the benefits and challenges of servitization, [email protected] manufacturing firms face a new imperative Laura Bahiense, COPPE/UFRJ, Brazil, decision to make. That is, if they would [email protected] implement services to their product-based portfolios, or not. Similarly, they would also Jean-David Caprace, LTS/COPPE, Brazil, need to identify the degree of services to [email protected] implement, which can vary between the Marcelo Igor De Souza, LTS/COPPE, basic helpdesks or maintenance services Brazil, [email protected] and the risk-sharing agreements. Before Eduardo Infante, UFSJ, Brazil, making these decisions, firms would want to [email protected] assess their organizations’ readiness for

60

Henrique Soares, SAGE/COPPE, Brazil, Solution (TOPSIS), Elimination and Choice [email protected] Expressing the Reality (ELECTRE), and Multi- Edilson De Arruda, SAGE/COPPE, Brazil, Attribute Utility Theory (MAUT). The strengths [email protected] and weaknesses of each models is discussed for the decommissioning activities Due to the maturity and short time span of after a presentation of available software’s. some economic activities, the demand for The review emphasized that AHP is one of decommissioning processes is rising steeply the most used methods in decommissioning over the last years. That rise is particularly studies. There is also various application of important in the oil and gas sector. Since outranking methods, such as PROMETHEE, such a sector often involves high complexity which aim to translate the preference and large amounts of investment, a careful relations established by decision makers. process is needed for a proper evaluation of However, it is clear that the Oil and Gas decommissioning alternatives. In addition, industry is not yet applying these since we are dealing with an economic methodologies and prefers simpler activity that affects many other sectors and techniques such as the comparative disciplines, a careful mapping of the assessment. Such a method employs a stakeholders is necessary and a holistic weighted sum of the evaluations with respect attention must be given to their needs and to each criterion with a view to select the concerns. Due to both the problem alternative with the best overall evaluation. It complexity and the need to involve multiple is a technique that is easy to apply and allows stakeholders, often with conflicting interests, compensation between the criteria, but the existence of a reliable and informative which may fail to integrate multiple tool to assist the decision-making process preferences. The outcome of the review becomes essential. Especially tailored for shows that the comparative assessment such problems, multi-criteria decision developed in several Oil and Gas industrial analysis (MCDA) methods can be a natural fit cases is probably not adequately for decommissioning problems. The aim of representing the complexity and the diversity this paper is to provide a review of the MCDA of technical and environmental situations. In methodologies that are commonly used for these terms, the development of a MCDA decommissioning activities in several decision-making methodology using economic segments with a focus on the oil integrating technical, environmental, life and gas sector. In this sector, cycle assessment, social, economic, safety, decommissioning processes date back to the risks, as well as regulatory, licensing and 1970s. The decommissioning process legal security issues is strongly generally takes place when the oil or gas field recommended. becomes uneconomical. It is often a time- consuming process. This is partly because it 4. Dimensionality Reduction may involve the partial or total removal of Approach for Multi-Criteria Problems: very complex structures, and partly because An Application to Oil and Gas Subsea it is subject to many regulations from Decommissioning different government bodies. An overview of decision support tools applied to *Isabelle Martins, SAGE/COPPE, Brazil, decommissioning in both academic papers [email protected] and industrial reports is presented in this Fernanda Moraes, SAGE/COPPE, Brazil, paper. The MCDA methods are not designed [email protected] to search for the best alternative with respect Giselle Tavora, LTS/COPPE, Brazil, to all criteria, the so-called ideal solution. [email protected] Instead, they seek compromises in real-world situations when there are conflicting criteria Eduardo Infante, UFSJ, Brazil, and no such alternative exists. Among the [email protected] available methods, one can find some Jean-David Caprace, LTS/COPPE, Brazil, standard approaches, such as Analytic [email protected] Hierarchy Process (AHP), Preference Ranking Marcelo Igor De Souza, LTS/COPPE, Organization Method (PROMETHEE), Simple Brazil, [email protected] Additive Weighting (SAW), Technique for Laura Bahiense, COPPE/UFRJ, Brazil, Order of Preference by Similarity to Ideal [email protected] 61

Edilson De Arruda, SAGE/COPPE, Brazil, of the most appropriate decommissioning [email protected] alternatives. The proposed approach makes use of machine learning techniques and The end-of-life of oil and gas structures has multi-criteria decision analysis (MCDA) tools become a worldwide concern. In many of the to generate groups of similar structures, for producing countries, decommissioning has which similar rankings are expected for the been largely discussed and several decommissioning alternatives. The similarity regulations and guidelines were generated is evaluated with respect to both the or are being elaborated. Among the diversity attributes of the structures and the of structures that are part of the offshore evaluations with respect to each criterion system, subsea installations demand a within a training set, with a view to generate special attention due to both their sensitive clusters for future application of the MCDA nature and the logistic challenges associated tools. We evaluated the predictive accuracy to the decommissioning process. In addition, of decision trees, random forests, gradient decommissioning is becoming more complex boosting machines and support vector as time lapses due to the increased number machines. After selecting a model, the of pieces of equipment operating in deep variables influence on the alternative waters. Initially, possible end-of-life selection was also analysed in order to measures includes in-situ abandonment and reduce the number of features required for total or partial removal. In addition, total or the study. Despite the arrival of the so-called partial removal can be achieved by means of big data era, some fields still lack massive different alternatives, which depend on the datasets and, unfortunately, available removal techniques. Given the decommissioning is among them. Hence, our distinct impacts and a possibly large number study makes use of bootstrapping and of criteria to consider, selecting a statistical analysis to generate synthetic decommissioning alternative becomes a datasets with a view to validate the proposed rather challenging problem. To address such techniques. To demonstrate the benefits of a problem, one must account for the great the approach, this study is applied to a study variety of technologies and materials on decommissioned pipelines in the North available, as well as the distinct Sea. The results are promising and suggest environmental and socio-economic that classification techniques can assist the conditions of each locality. In order to decision making process with little impact to evaluate the alternatives, a multidisciplinary the overall accuracy of the MCDA technique. approach is needed which involves Preliminary results indicate better economic, environmental, technical, social performance for gradient boosting machines and safety aspects, among others. with mean accuracy of 82%. Furthermore, the possibly large number of stakeholders and their potentially conflicting MON-4-D interests can result in a very controversial process. The majority of published Tutorial: Improve Your Decisions by decommissioning processes so far have Learning and Experiencing the ANP relied on a methodology called comparative Best Practices- Part II assessment, often based on subjective Monday 16:00-17:40 - Room: Topkapı judgments by stakeholders with respect to a Palace number of pre-selected criteria. Generally, the alternatives are ranked by means of Chair: Birsen Karpak weighted sum of the evaluations with regards to the selected criteria. However, the Improve Your Decisions by Learning process must be applied to each individual and Experiencing the ANP Best piece of equipment, in a process that can Practices- Part 2 of 2 become rather cumbersome for large offshore systems. In this work, we propose a *Orrin Cooper, University of Memphis, classification technique to define groups of United States, [email protected] similar pieces of equipment for which a joint For details see Section MON-3-D analysis would be applied. Therefore, this paper is focused on reducing the problem Please bring your own laptop to this special dimension, while also ensuring the selection session

62

MON-4-E approximation to the whole Pareto set for a given approximation error. In general, these Invited Session: Solution Methods for algorithms solve a Pascoletti-Serafini (P-S) Various Multiobjective Optimization scalarization model in each iteration. P-S Problems scalarization has two parameters: a reference point that is chosen from the Monday 16:00-17:40 - Room: Hagia Sophia objective space and a direction vector Chairs: Özlem Karsu, Firdevs Ulus chosen from the ordering cone. The choice of these parameters affects the overall 1. Representation with Respect to efficiency of the algorithm. In the original Hypervolume: A Scalarization algorithm the direction parameter is fixed Method for Bi-Objective Optimization and the reference point is taken as one of the Problems vertices of the current outer polyhedron. We propose different variants of the algorithm *Britta Schulze, University of Wuppertal, where the direction parameter changes in Germany, [email protected] each iteration and different selection wuppertal.de procedures are applied for choosing the reference point. We compare the efficiencies The hypervolume indicator is a quality of the variants through numerical analysis. measure for representations of the nondominated set of multi-objective 3. An Improved Hyperboxing optimization problems. It measures the Algorithm For Multi-Objective Non- hypervolume of the region that is spanned by Linear Optimization Problems the points of the representation and an a priori given reference point. Its main *Kerstin Daechert, Fraunhofer ITWM, drawback lies in its computational cost. Germany, Therefore, approximation algroithms have [email protected] been proposed that reduce the Katrin Teichert, Fraunhofer ITWM, computational effort. While the hypervolume Germany, is heavily used in the context of evolutionary [email protected] multi-objective optimization algorithms to evaluate the fitness of solution candidates, A popular way to solve a continuous multi- its application as quality measure in objective optimization problem is to compute combination with exact methods is rather an inner and an outer approximation to unusual. In this talk we present a sandwich the Pareto front in between. While scalarization method for bi-objective particular sandwich algorithms exist for optimization problems that allows to build a convex problems, we can also define a very representation, with a quality guarantee with general sandwich algorithm in the non- respect to hypervolume, in an apriori convex case by only using the properties of manner. Theoretical properties and non-dominance. Since the part of the algorithmic aspects are discussed in the outcome set that is sandwiched between the context of selected multi-objective inner and outer approximation can be combinatorial optimization problems. represented as a union of rectangular sets, we call this general sandwich approach 2. On the Efficiency of Benson-Type hyperboxing algorithm. We base our Approximation Algorithms for Convex hyperboxing algorithm on previous work Multiobjective Optimization presented in Teichert (2014). Therein, it was Problems already shown that the algorithm results in a very good coverage of the Pareto front. *Firdevs Ulus, Bilkent University, However, the computational burden of the Turkey, [email protected] algorithm was shown to be very high which İrem Nur Keskin, Bilkent University, made the algorithm impractical for Turkey, [email protected] dimensions higher than five. In this talk we show how to incorporate recent ideas from Benson-type approximation algorithms for the discrete context to speed up the convex multiobjective optimization problems algorithm. More precisely, we use improved generate an inner and an outer techniques to update the inner

63 approximation in the discrete context. These MON-4-F techniques can be translated in a straightforward way to the continuous case. Contributed Session: Multiobjective We use the improved update of the inner Metaheuristics for Challenging and, analogously, outer approximation within Applications the hyperboxing algorithm. We present the main ingredients of the hyperboxing Monday 16:00-17:40 - Room: Basilica algorithm and show improvements with Cistern respect to computational time compared to Chair: Alfonso Mateos Teichert (2014). Our numerical results are based on test problems with up to 10 1. Multi-Objective Grouped Genetic objectives. We show that the algorithm Algorithm for the Joint Order results in a very good coverage at reasonable Batching, Batch Assignment and times for a set of various test problems. Sequencing Problems 4. Solution Approaches for Equitable *Jose Cano, Universidad de Medellín, Multiobjective Programming Colombia, [email protected] Problems Pablo Cortés, Universidad de Sevilla, *Özlem Karsu, Bilkent University, Spain, [email protected] Turkey, [email protected] Emiro Campo, ESACS - Escuela Superior Bashir Bashir, Bilkent University, en Administración de Cadena de Turkey, [email protected] Suministro, Colombia, [email protected] We consider multi-objective optimization Alexander Correa-Espinal, Universidad problems (MOP) where the decision maker Nacional de Colombia, Colombia, (DM) has equity concerns. We assume that [email protected] the preference model of the DM satisfi es properties related to inequity-aversion, This article aims to solve the joint order hence we focus on finding nondominated batching, batch assignment and sequencing solutions in line with the properties of problem (JOBASP) considering multiple inequity-averse preferences, namely the picking devices in warehouse systems. The equitably efficient solutions. We discuss two JOBASP groups several customer orders in a algorithms for finding good subsets of single picking list or batch, as well as assigns equitably efficient solutions. The fi rst batches to the multiple picking devices approach is an extension of an interactive according to their load capacities and approach developed for finding the most schedules the batches assigned to each preferred nondominated solution when the picking device in order to achieve a high utility function is assumed to be picking efficiency and minimize the quasiconcave. We fi nd the most preferred tardiness. To solve the JOBASP, we propose equitably efficient solution when the utility a multi-objective grouped genetic algorithm function is assumed to be symmetric (GGA) to minimize total travel time and total quasiconcave. In the second approach we tardiness. The proposed encoding scheme is generate an evenly distributed subset of the composed of three levels: picking devices, set of equitably efficient solutions to be number of the batch, and customer orders considered further by the DM. We show the assigned to each batch. An order group computational feasibility of the two procedure is used to create the initial algorithms on equitable multi-objective population. It uses an order pool to place knapsack problems that fund projects in orders that have not been assigned yet to a different categories subject to a limited batch. This procedure chooses randomly a budget. We perform experiments to show picking device and assigns it to a gene, then and discuss the performances of the customer orders are randomly chosen from algorithms. the order pool and assigned to each batch. The fitness function is defined as the weighted sum of the total travel time and total tardiness. The selection of parental chromosomes for the crossover operator is

64 based on the linear selection ranking Naoyuki Shibukawa, Hiroshima method. Pairs of solutions are chosen as University, Japan, b155750@hiroshima- parents using the roulette wheel selection. u.ac.jp Two crossing points delimit the crossing Ichiro Nishizaki, Hiroshima University, section to be exchanged between each pair Japan, [email protected] of parents. A correction mechanism is applied to fix infeasible offspring. Orders that Tomohiro Hayashida, Hiroshima have not yet been assigned to a batch University, Japan, complete each chromosome by means of the [email protected] order group procedure. The survival Yuuki Kanazawa, Chubu Electric Power mechanism ensures that individuals with Co. Inc., Japan, better performance prevail in each [email protected] generation. Immigration defines the number Hirokazu Uenishi, Chubu Electric Power of new individuals to create using the order Co. Inc., Japan, [email protected] group procedure. In the mutation operator, two genes are randomly selected and then This research proposes an approach for deleted, and new genes are created using handling electricity distribution system asset the order group procedure. The genetic planning that is integrated with many algorithm will stop when the number of objective reconfiguration problem of the generations is satisfied. We produced network topology. The distribution systems several test instances based on parameters supplying the electricity to consumers such as customer orders, warehouse include various important assets such as configurations, and weighing parameters. distribution lines, tap-changing transformers, The combination of experimental parameters and switchgears. Many remote and manual makes 27, and we have generated 10 switchgears are located on the distribution instances for each scenario, providing 270 lines to keep supplying the electricity to test problems. The performance of the GGA consumers even in case of faults and is compared to adapted heuristics based on maintenance of the assets. Since the the FCFS and EDD rules respectively. The network topology is radial to avoid the failure proposed GGA outperforms the FCFS rule- extension, there is a single route of the based heuristic providing average savings of electricity between the substation and 25.4%, and maximum savings up to 46%. consumers in a normal state. By opening and Likewise, the proposed GGA outperforms the closing the switchgears in a proper order, EDD rule-based heuristic providing average alternative routes of electricity can be made savings of 10.2%, and maximum savings up flexibly in the emergency conditions. The to 28%. Likewise, computational reconfiguration problem by controlling the experiments provide an average of 54.2 switchgears has many objectives to be seconds for each scenario. As a conclusion, minimized: (#1) capacity of consumers with the proposed GGA provides important power outage, (#2) total travel distance of savings on travel time and tardiness. The technicians to control the manual implementation of the GGA is feasible in real switchgears on site, (#3) the total length of warehouse environments due to the time to control the switchgears, (#4) the computing times obtained in the operation times of switchgears, and (#5) the computational experiments. operation times of tap-changing transformers. There are trade-off 2. A Many Objective Optimization relationships between these objectives. Approach for Distribution System Although the objective #1 is the highest Asset Planning Integrated with priority needless to say, the other objectives Network Reconfiguration are also considered to be important for the distribution system management. This *Shinya Sekizaki, Hiroshima University, problem is a many objective optimization Japan, [email protected] problem. If there are n switchgears with two Takuya Yamasaki, Hiroshima University, states (opened or closed) in the distribution Japan, [email protected] system, total number of network topology including infeasible solutions is 2n. Since the reconfiguration problem has heavy computational burden, many-objective

65 evolutionary algorithms (MaOEAs) are 3. A Real Time Adaptation of the ATC effective to find non-dominated solutions Work Shift Scheduling Problem to approximating the true Pareto front. We have developed the MaOEA based on NSGA-III in Deal with Incidents in Airports our previous research to solve the many- Control Centers objective optimization problem for the large- *Alfonso Mateos, Universidad scale distribution system within the practical Politécnica de Madrid, Spain, computational time. The distribution lines have the reserve capacity for redundancy. [email protected] Due to the redundancy, the distribution Antonio Jiménez-Martín, Universidad system operator can choose a Politécnica de Madrid, Spain, reconfiguration procedure among [email protected] candidates. In this research, we integrate the Faustino Tello, Universidad Politécnica reconfiguration problem with curtailment of de Madrid, Spain, [email protected] the redundant line capacity because the curtailment can reduce the cost of money, if A very important problem in air traffic control possible. In our proposed method, a set of (ATC) is to determine the minimum number the non-dominated solutions Si as the of air traffic control operators (ATCo) reconfiguration procedures for maintenance necessary to cover an airspace sector of the asset i is computed by the developed configuration, while satisfying certain ATCo MaOEA. Based on a set of the line current of labor conditions, including, for instance, line j in case of the reconfiguration for resting and working time distributions. maintenance cases i = 1, 2, …, L, the Alternatively, the number of ATCos could be curtailable capacity of line j and fixed. The aim then would be to distribute corresponding objective values are ATCos to cover the corresponding airspace computed. If the distribution system operator sector configuration. This problem is known does not satisfy the objective values, he/she as the ATCo work shift scheduling problem. can specify upper bounds of the objective A great number of constraints accounting for values. After the specification of the upper ATC labor conditions have to be taken into bounds, the proposed method provides account in that optimization problem. In him/her with alternative curtailable capacity Spain, all these conditions were compiled of line j satisfying the constraints. In order to and published in the Official State Gazette validate the proposed method, realistic (Boletín Oficial del Estado, BOE), Royal computational experiment is performed Decree 1001/2010, and Law 9/2010, using a large-scale distribution system model regulating the provision of air traffic services. of Chubu Electric Power Co. Inc. in Japan. These labor constraints refer to ATCos This model has 142 switchgears (remote: minimum and maximum resting and working 29, manual: 113), three feeders, 16 tap- periods, number of sector changes, times in changing transformers, (L =) 193 nodes, and the same sector… Besides, a multi-objective 200 lines. The above five objectives are perspective is incorporated into the problem considered in the experiment. The developed to account for different conflicting objectives. MaOEA gives more than 2,000 non- First, the ATCo work and rest periods and dominated solutions to us within 30 h. The positions should be as close as possible to line current flowing through line j = 1, 2, …, fixed values. Second, the solution structure 200 is computed based on the obtained sets should be similar to the previous template- of non-dominated solutions Si, i = 1, 2, …, based solution. This makes the solution 193. The proposed method successfully more understandable for the control center compute the curtailable line capacity by staff and would facilitate any possible hand- reconfiguration of the network topology. made changes. Additionally, the number of control center changes should be minimized. Finally, the ATCo workload distribution should be balanced. A three-phase problem- solving methodology was proposed to deal with such problem. In the first phase, a heuristic is used to derive infeasible initial solutions on the basis of templates. Then, a multiple independent run of a metaheuristic

66 is conducted aimed at reaching feasible 4. A Hybrid Algorithm to Solve the solutions in the second phase. Finally, in the Multi-Objective Spanning Tree third phase, a multiple independent metaheuristic run is again conducted from Problem the initial feasible solutions to optimize the *Asma Boumesbah, USTHB, Algeria, objective functions. To do this, we account [email protected] for ordinal information about the relative Mohamed El-Amine Chergui, USTHB, importance of objectives and transform the Algeria, [email protected] multiple to a single optimization problem by using the rank-order centroid function. The The multi-objective minimum spanning tree performance of using simulated annealing problem (MOST ) has been shown to be NP - (SA) and variable neighborhood search (VNS) hard even with two criteria. We propose an has been analyzed on the basis of four algorithm that exploits the advantages of representative and complex instances of the both ”Non-Sorting Genetic Algorithm” (NSGA- problem corresponding to different airspace II) and “Variable Neighborhood Search” sectorings provided by the Spanish ATM (VNS) metaheuristics to find a good Research, Development and Innovation approximation to the Pareto front of the Reference Center (CRIDA). We conclude that MOST problem. Taking into account the both metaheuristics have a similar specificity of the problem, this algorithm first performance but VNS is faster than SA. In adopts novel two point crossover operator this paper, we consider a variant of the above and a heuristic mutation operator in NSGA-II ATC work shift scheduling problem. We to obtain the first approximation of the assume that a solution derived from the Pareto front, front 1. In the second phase, above methodology is being used at any the VNS based algorithm is called for further moment at the Control Center in an airport, improvements of front 1. It allows finding i.e. an airspace sector configuration has good individuals to counterbalance between been identified and ATCos have been the diversification and the intensification assigned to position and are currently during the optimization search process. working (or resting). Now, we assume that an Experimental comparison studies show to incident happens. For instance, one several promising results and indicate that the ATCos get sick and can not do their job, a proposed algorithm does not have the nearby airport may close for meteorological limitations of previous algorithms because of reasons and the traffic is redirected to our its speed, its scalability to solve MOST airport… In such situations, the current problem considering complete graphs with airspace sector configuration should be more than 200 nodes, and its ability to find changed from that instant to the end of the both the supported and non-supported work shift and, consequently the ATCos must solutions. be reassigned accordingly to the new configuration but taking into account the part of the working shift they have already made. In this new problem, the response time is crucial since planes are in the air and must be controlled and sometimes no feasible solutions are possible, leading to the sacrifice of some ATCo labor conditions. We have again compared the performance of SA and VNS to dealt with this variant of the ATC work shift scheduling problem, aimed at achieving the less unfeasible solution in few seconds. For this, the experts from CRIDA have provided ordinal information about the ATCo labor conditions.

67

2. Application of Multi-Criteria Decision-Making Methods in TUESDAY Healthcare Management: A Comprehensive Literature Review TUE-1-A *Renata Pelissari, UNICAMP, Brazil, Invited Session: MCDM/A for [email protected] Assessment Issues in Healthcare and LCA Sarah Ben Amor, University of Ottawa, Tuesday 9:00-10:40 - Room: Galata Tower Canada, [email protected] Chair: Sarah Ben Amor Healthcare involves several complex decision-making situations, from decisions 1. Decision Framework for Drug directly related to medical treatments to Formulary Design Using Multiple those regarding public-health policies and Criteria Decision Analysis financial reimbursements. Multi-criteria decision making (MCDM) tools are *Gilles Reinhardt, Université d'Ottawa, increasingly used in the healthcare sector. Canada, [email protected] The purpose of this paper is to conduct a Sarah Ben Amor, University of Ottawa, systematic literature review on the Canada, [email protected] application of MCDM techniques in Vusal Babashov, Université d'Ottawa, healthcare over the last twenty years. About Canada, [email protected] 130 papers were selected from Scopus for analysis. Those papers, published in more Reviewing new drugs to determine coverage than 80 international journals between 1998 or reimbursement level is a complex process and 2018, were classified according to the that involves significant time and expertise. following criteria: method used for Review boards gather evidence from the preference modeling, method used for provided dossier, input from clinicians and aggregation, source of the data used in the patients, and results of clinical and economic modeling, uncertainty modeling, sensitivity reviews. This information consists of analysis, participatory approach/group assessments of the drugs on multiple criteria decision, hierarchical criteria, interaction that often conflict with one another. We use among criteria, context of application, a recent extension of the UTilitiés Additives publication year, scientific journal and DIScriminantes approach, UTADIS-GMS (the country of case-studies. The results show suffix refers to its authors) which is based on that in healthcare, group decision-making the preference disaggregation paradigm. dominates over single-person decision The UTADIS approach has the ability to making: about 80% of the papers selected deconstruct a portfolio of elements such as considered multiple decision-makers. Unlike a set of drugs that have been reviewed and sectors in which the application of MCDM decisioned. It derives criteria weights that methods involves the evaluation of one or a are consistent with the preferences exhibited few decision-makers (such as supplier by the review boards in their selection, sustainable manufacturing and recommendations. The GMS variant of the portfolio project selection), in healthcare, UTADIS approach provides a flexible decision almost half of the papers conducted surveys framework that considers various forms of whose results became the source of the data preferences for the decision making bodies used in the modeling. In addition to the fact and yields richer information in terms of both that most papers considered group necessary and possible assignments. It also decisions, we observed that in healthcare, it offers more flexibility notably because it is important to consider several opinions accepts imprecise assignments. We show from different experts instead of focusing on how to adapt the methodology to the context one or a few assessments. The cases that of formulary decisions and apply it to considered the preferences of only one oncology drugs reviewed in Canada between decision-maker are typical of situations in 2011 and 2017. We also illustrate how to which the decision concerns only one patient conduct scenario analyses and how the such as using an app to select a physician) method can be applied for predictive and not to decisions that may impact purposes. multiple patients (such as public policy or 68 patient prioritization in a hospital). The main Sarah Ben Amor, University of Ottawa, contexts of MCDM application are clinical Canada, [email protected] decision-making, including diagnoses and Yvan Beauregard, École de Technologie treatments, and healthcare waste Supérieure, Canada, management followed by health technology [email protected] assessment. Other application contexts are public health and policy, site selection for Marc Paquet, École de Technologie healthcare facilities, reimbursement Supérieure, Canada, decisions, risk assessment, decisions linked [email protected] to professional practices, resource allocation Multicriteria decision making tools in and schedule planning. AHP was identified technological obsolescence management as by far the most frequently applied method are becoming crucial in the current highly for preference modeling. It is also the main competitive market. Parts obsolescence is a method used for aggregation, with TOPSIS complex problem involving many factors and and its fuzzy versions also widely applied. stakeholders. Its impact on the Almost half of the papers considered manufacturing system can be minor uncertainty in the decision process, mainly (obsolete part not critical), medium (medium due to the vagueness inherent in the criticality) or high (high criticality). Managing linguistic terms used by the experts involved parts obsolescence is a multi-criteria in the decision-making process. To model decision making problem in which different uncertainty, fuzzy-sets theory was the main strategies such as last time buy, redesign or technique applied. We concluded that salvage are assessed in regard to different MCDM can provide a more comprehensive criteria such as cost, life cycle assessment, and transparent approach in healthcare to environmental impact or obsolete parts systematically capture patient, expert and characteristics. Obsolete parts stakeholder concerns, to elicit their value characteristics are an important factor in the preferences, to assess various criteria and to risk impact scale and management. reconcile all competing interests and trade- Considering the wide variety of obsolete offs. We also identified these abilities as the parts and their extremely variable main reasons for the growing application of characteristics, classical weighting methods MCDM in the healthcare context. The seem unable to handle the model’s criteria methods generally applied in healthcare do weight dynamic adjustments with respect to not take full advantage of recent the manufacturing system performances. In developments in the MCDM field. Methods this paper, an agent-based method is such as AHP have limitations which new developed to assess, dynamically, the methods have overcome. Therefore, in future criteria weights based on the parts’ research, decision-making in healthcare may specifications. However, it is difficult to integrate the application of new MCDM directly compare the proposed algorithm techniques that represent the decision performance with existing methods in the process in a more realistic way and lead to literature, as there is no benchmark for more robust results. These techniques may dynamic criteria weighting in obsolescence include volume-based methods to elicit management. An illustrative example is preference information (e.g. SMAA), methods presented based on a static example from that allow for the modeling of hierarchical the literature. The results are then compared criteria (e.g., MCHP), methods that consider for validation. multiple types of uncertainty concurrently or methods that consider interaction among 4. The Integration of Multicriteria criteria (e.g. Choquet Integral). Decision Aiding with Life Cycle 3. Agent-Based Algorithm for Assessment from Methodology to Dynamic Criteria Weight Assessment Practice – A Critical Review in Multi-Criteria Decision Making for *Sarah Ben Amor, University of Ottawa, Obsolescence Management Canada, [email protected] *Imen Zaabar, École de Technologie João Clímaco, INESC- Coimbra, Supérieure, Canada, Universidade de Coimbra, Portugal, [email protected] [email protected]

69

In this work we provide a critical review on produces errors. Moreover, it must be noted the integration of Multicriteria Decision that the direct use of weights involves a Aiding (MCDA) with Life Cycle Assessment normalization of the terms, which may also (LCA) and Life Cycle Sustainability contribute to the distortion of the results. Assessment (LCSA). LCA/LCSA are proposed Multi-Attribute Utility Theory (MAUT) can help both to characterize present situations, and in eliciting the weights in a more rigorous to recommend future policies. Two questions way, but it does not overcome the root of the should be raised: (1) Is LCA/LCSA useful in problem. Some authors have proposed decision aiding? (2) Can MCDA help in avoiding the full compensatory approaches making compatible LCA/LCSA and decision by using outranking methods based on aid science, in order to obtain fruitful partial order relations, such as those of the recommendations? This review involves a ELECTRE family, followed by some kind of classification of the literature and a aggregation and exploitation of the results. discussion on the proposals and/or of the Although avoiding the complete results. Different types of life cycle concepts compensation, these methods involve fixing and issues are considered — from the well several parameters and changes in these characterized classical LCA, on which ISO parameters can drastically influence the norms were formalized in 2006 to help in the results. Finally, simple non-compensatory implementation of a tool dedicated to the procedures have also been used in the environmental impact assessment (LCIA), to aggregation of outputs in LCA/LCSA. Inter- the Life Cycle Sustainability Assessment criteria aggregation is particularly sensitive in involving two other dimensions: Social Life LCSA, as data is highly diverse in nature and Cycle Assessment (SLCA) and Life Cycle can be either quantitative or qualitative, Costing (LCC). Note that, SLCA introduces involving uncertainty and lack of information. subjective and complex modelling issues As subjects deserving of further research we which are not yet well regulated and which highlight the following: comparison of the use depend on epistemological assumptions, of different MCDA and holistic approaches in and that the use of LCC is far from life cycle evaluation of products and services, consensual. We also analyse the relevance namely dealing with uncertainties; sensitivity in literature of some complementary analysis, robustness analysis and concepts and related approaches, namely: visualization tools; association of system Life Cycle Thinking, Life Cycle Management, tools with MCDA; combination of Consequential Life Cycle Assessment and participatory approaches with MCDA; Dynamic vs Static Life Cycle Assessment. We development of MCDA group decision noticed that decision aiding structuring approaches; using MCDA to compare LCA techniques can be used in the first phases of with other approaches regarding the LCA process, namely: in the definition of environmental evaluation and exploiting the objectives and in the Life Cycle Inventory, connections among LCSA, MCDA and taking into account challenging uncertainties Circular Economy. and lack of information. The characterization and modelling of the process seem crucial to TUE-1-B evaluating the success of LCA/LCSA in Contributed Session: AHP/ANP practice, particularly with regards to study Applications in Academia and Education goals and boundaries, data collection and management, the definition of impacts, as Tuesday 9:00-10:40 - Room: Maiden's Tower well as indicators/criteria and alternatives. In Chair: Monica Garcia-Melon addition, a large number of papers were dedicated to the aggregation of the 1. Selection of Part-Time Employed multidimensional output of the LCA/LCSA Students with MCDM Approach: A processes. Although in general multi- attribute approaches are adequate, in some Case Study in KDPU situations multi-objective methods were *Büşra Günhan, Kütahya Dumlupınar used. Typically, the aggregation of outputs is University, Turkey, carried out using weighted sums of [email protected] normalized outputs. Many recent works have cautioned that the fully compensatory use of In this study, an analytical approach is weights raises several questions and/or proposed to select the part time employed

70 students for various departments of the created for AHP process was solved by using university. In the first step, criteria are packaged software Expert Choice 11. In this determined by regarding the previous paper, an MCDM methodology is used experience of the commission members of consisting of AHP and SMART techniques for Metallurgical and Materials Engineering part-time employed student selection. The Department, the procedures and principles obtained results show that AHP methodology which are listed in the Administration can be more useful to determine the best Guideline and expected qualifications stated candidate for part-time employed student in the Application Form by Kütahya because of its providing aggregation of Dumlupınar University (KDPU). Secondly, the individual priorities. evaluation of the alternatives has been analysed by two different multi-criteria 2. Using Analytic Hierarchy Process, decision making (MCDM) method by taking Consistent Fuzzy Preference into account the experience of the decision- Relations and Best Worst Method to makers who work as lecturers. The results Construct Peer Assessment Method obtained by analytic hierarchy process (AHP) and simple multi-attribute rating technique Xia Li, Sanming University, China, (SMART) have been compared. With the [email protected] part-time employment of students in various Jin-Hua Luo, Sanming University, China, departments of KDPU, such as engineering [email protected] faculty and the (centre) library etc., it is Gwo-Hshiung Tzeng, Graduate Institute aimed to help students contribute to their of Urban Planning, Taiwan, individual skills, like productivity and labour [email protected] discipline, by employing them in accordance with their interests and skills. It is detected *Shan-Lin Huang, Sanming University, that the current procedure is performed by China, [email protected] using a number of intuitive and subjective Group report is a form that teachers often assessment processes by departmental use in school teaching, but teachers usually commission members. This procedure is only score the group and it is difficult to rate insufficient to achieve fair, sustainable, and individuals. The method used in the past was peer-conflict preventive outcomes among to convert the group's grades into individual students who belong to different socio- scores through the contributions of the economic background, and are in the same students. The purpose of the AHP (Analytic success level. From this point of view, it is Hierarchy Process) method is to obtain the aimed to have objective evaluation for the sequence of the elements (refers to students part-time employed student selection by in the group) through pairwise comparison. using multi-criteria decision making. In the Therefore, some scholars have proposed to literature, there are a lot of studies related to use AHP to solve this problem. However, the personnel selection in several industry and main task of education is to learn, not to student selection problems for changing “give points”. Therefore, AHP still has purposes. Yeh (2003) formulated a problems with too many items. Consistent scholarship student selection problem as an fuzzy preference relationships can also MADM problem and proposed a MAVT-based identify the order of elements, but the MADM method as a solution. Afshari et. al. number of items can be greatly reduced. (2010) studied on personnel selection Therefore, some scholars have refined the problem by using simple additive weighting measurement and calculation methods of (SAW) approach. Khorami and Ehsani (2015) peer assessment with this method. BWM reviewed the main MCDM approaches for (Best Worst Method) is another technique to personnel selection problems. Deliktas and get the order of elements with fewer Ustun (2017) proposed a fuzzy multiple problems, but no one has used this method criteria decision making approach to select to explore it in the past. Therefore, the an industrial engineer in a manufacturing purpose of this paper is to use these three environment. For SMART and AHP technique, methods to improve the peer assessment the weights of criteria was determined by method, so that teachers can judge the using brainstorming method. SMART contribution of students in a simpler but technique was applied by using MS Excel more accurate way. templates and the hierarchical structure

71

3. Knowledge Sharing Limitations among the other influencing group of among Academia: Analytic Network barriers which are 23.5%, 22.4%, 21.2% for unhealthy rivalry among institution units, Process Approach loose and weak institutional structure, and *Olamilekan Oyenuga, University of lack of knowledge strategies respectively. Lagos, Nigeria, This study provides assistances to policy [email protected] maker in formulating and implementing effective knowledge sharing strategies that Sulaimon Adebiyi, University of Lagos, can reverse ugly influences of organizational Nigeria, [email protected] and individual factors on knowledge co- Olamitunji Dakare, Pan-Atlantic creation in order to improve the productivity University, Nigeria, of academics towards actualizing university [email protected] missions. Thereby, it proffers insight into the Charles Omoera, University of Lagos, extent which organisational and individual Nigeria, [email protected] factors impede knowledge sharing in the process of actualizing the university The purpose of this paper is to estimate the missions. Keywords: Knowledge sharing, relative influences and implications of Knowledge sharing limitations, Analytic knowledge sharing limitations among network process, Academia. academia on the university missions (Academic-industrial research, quality of 4. Indicators for Monitoring teaching and entrepreneurship Responsible Research and engagement). The study aims at providing an Innovation in Spain: The Case of understanding of application of multi-criteria decision making analysis through Analytic Science Education Network process (ANP) for the evaluation of *Monica Garcia-Melon, Universitat knowledge sharing limitations among Politecnica de Valencia, Spain, academia showing complete ranking with [email protected] scores. The study employed a quantitative approach in line with positivism tradition. Hannia González-Urango, Universitat Data were collected through an ANP-based Politecnica de Valencia, Spain, self-administered questionnaire to academia [email protected] on the pairwise comparison of clusters in the Blanca Pérez-Gladish, Universidad de network (individual and organisational Oviedo, Spain, [email protected] factors) and nodes (university missions). Maria Regla Vinagre-Fernandez, Sample for the study were drawn through a Universitat Politecnicia de Valencia, three-stage multi-stage sampling procedures Spain, [email protected] in order to capture different academic discipline across faculties and levels. Data This research seeks to contribute to the generated were analysed through ANP proposal of a framework for the development procedure using super decision software for of indicators of Science Education through generating the super matrix (unweighted, involving in the decision-making processes weighted, and limit) which finally yield the discussions among the different agents of relative priority of components within the the scientific and innovation systems in network system. The study reveals that Spain. The research is frame-worked in the components of the individual barriers current promotion of the term Responsible (cultural difference, fear of loss of knowledge Research and Innovation (RRI) in Europe and power, knowledge hoarding culture, lack of its impact on national scientific and communication skills, and lack of trust and innovation policies. As a result, two time) have almost equal influences on European-based sets of indicators for the six knowledge sharing limitations while strong areas related with RRI (governance, public depth of politics within the work place was engagement, gender equality, science moderately perceived as the most education, open science/access and ethics) influencing hindrance to knowledge sharing have been proposed. To facilitate the among academia with one-third (33%) monitoring of these areas, the selection of among other organizational barrier factors. smaller set of indicators adapted to the This was followed by influential differences reality of each R&D context has been

72 suggested (Expert Group on Policy Indicators Spanish R&D+i systems, AHP questionnaires for Responsible Research and Innovation, to pairwise compare the relevance of such 2015). This study aims to propose and test indicators will be answered by them. With the a decision model to weight the most relevant information obtained from these indicators for public engagement to monitor questionnaires, the more relevant indicators national R&D initiatives using the case of for this specific group of stakeholders will be Spain as an example. In the case of Spain, identified. The application of such a the recent published National Plan for technique, will allow policy-makers and Scientific and Technical Research and decision-makers to identify the more Innovation (2017-2020) includes in its fifth relevant indicators for a specific context and objective “Promotion of an open and select reduced sets of indicators to gather responsible model of R&D+I supported by information on Science Education the participation of the Society” references to performance. open access and open science, ethics, gender and public engagement. Science TUE-1-C Education (SE), as one of the key areas Contributed Session: Preferences, proposed by the EC, refers to both: - better Comparisons and Uncertainties equip future researchers and other societal actors with the necessary knowledge and Tuesday 9:00-10:40 - Room: Dolmabahçe tools to fully participate and take Palace responsibility in the research and innovation Chair: Matteo Brunelli process and, - boost the interest of children and youth in maths, science and technology 1. Multiple Criteria and Social so they can contribute the researchers of Influence Analysis for Voluntary tomorrow and contribute to a science-literate society. This work is based on three Vaccination Decision Making hypotheses: 1. Indicators to monitor and Yu-Wang Chen, The University of evaluate SE initiatives and policies might not Manchester, United Kingdom, have the same relevance in different [email protected] contexts. 2. Indicators to monitor and evaluate SE initiatives and policies can be The achievement of widespread immunity to prioritized. 3. The AHP methodology might be infectious diseases like influenza by a useful tool to propose a strategy to adapt vaccination is very important for public and select the more relevant context-based health. Thus, it is essential to understand the sets of indicators to specific contexts that drivers that contribute to voluntary vaccine allow stakeholders to propose new uptake for the purpose of developing indicators. The methodology proposed for interventions to increase vaccine coverage. our work is based on three steps: 1. Previous studies have shown that there are Identification of the most relevant actors in multiple factors affecting an individual’s SE (Social Network Analysis) 2. Generation of intention to vaccinate, including safety of a list of context-based SE indicators (Experts vaccines, severity of diseases, potential side panel) 3. Prioritization of the list of indicators effects and so on. However, limited studies proposed (AHP) We rely on the opinion of have considered how individuals’ perception, experts and stakeholders in the area of judgement and decision on voluntary Science Education in Spain to provide vaccination may update due to the spread of different weights for the list of agreed social influence. In this work, an integrated indicators. The panel of experts will include model is proposed to analyse the dynamics representatives with competences in science of social influenced voluntary vaccination and innovation policies from the local, decision making, which takes into account regional and national governments, both the formulation of individuals’ academics with background in research on subjective judgements and the science education and with practical characterisation of influence spreading in a experience in its implementation (science social network. It is assumed that individuals museums, science centres, schools, in social networks can explicitly and academies, science magazines, science independently express their perceptions of blogs and living labs). Once the experts have vaccination. In the meantime, those who fail agreed a set of relevant SE indicators for the to make a firm decision initially can be

73 influenced by their social neighbours. As a plurality of mindsets. With this in mind, we result, individuals’ subjective judgements on propose a novel methodology taking an vaccine uptake are characterised by belief approach similar to Stochastic Multi-criteria distributions in the context of multiple Acceptability Analysis. Considering all the criteria decision analysis (MCDA), whilst rankings of alternatives corresponding to the social influence is reflected in the form of different mindsets, our methodology belief spreading and updating against each estimates the probability that an alternative of the decision criteria. The evidential attains a given ranking position together with reasoning (ER) rule, as a further the probability that an alternative is development of the seminal Dempster- preferred to another one. Since the Shafer (D-S) theory of evidence, is then exponential number of spanning trees applied to aggregate multiple pieces of makes their enumeration prohibitive, we uncertain information resulted from both propose computing approximate subjective judgements and social influence. probabilities using statistical sampling of A series of experimental simulations are decision trees. Our approach can also be conducted to analyse the dynamics of applied to incomplete sets of pairwise voluntary vaccination decision making under comparisons. We demonstrate the social influence in a social network. The methodology's usefulness with the famous effects of information flow and the sensitivity example of the school selection problem. We of individuals’ preference towards certain estimate the probabilities of each school criteria are explored in a systematic manner. attaining a given ranking position together This exploratory work provides a novel with the probabilities of one school preferred perspective to analyse social influenced over other. The usefulness of statistical decision making, and the research findings sampling is further demonstrated with the can be useful for disease control help of another real-life policy making issue departments to better understand the involving the selection of telecom networks dynamics embedded in an individual’s in rural areas. vaccination decision as well as observe the trend of vaccinating behaviours in a social 3. SURE: A Method for Decision- community. Making under Uncertainty 2. Supporting Decisions by *Richard Hodgett, University of Leeds, Unleashing Multiple Mindsets Using United Kingdom, Pairwise Comparisons [email protected] *Sajid Siraj, Leeds University Business Managerial decision-making often involves the consideration of multiple criteria with School, United Kingdom, high levels of uncertainty. Multi-attribute [email protected] utility theory, a primary method proposed for Salvatore Greco, University of Catania, decision-making under uncertainty, has been Italy, [email protected] repeatedly shown to be difficult to use in Michele Lundy, Dublin City University, practice as it specifies uncertain outcomes Ireland, [email protected] by means of probability distrubutions which are not typically known. This talk presents a Inconsistency in pairwise comparison novel approach termed Simulated judgements is often perceived as an Uncertainty Range Evaluations (SURE) (DOI: unwanted phenomenon and researchers 10.1016/j.eswa.2018.08.048) which helps have proposed a number of techniques to decision makers in the presence of high either reduce it or to correct it. We consider levels of uncertainty. SURE has evolved from Spanning Trees Analysis which is a recently an existing method that has been applied emerging idea for use with the pairwise extensively in the pharmaceutical and comparison approach and which represents speciality chemical sectors involving the plurality of mindsets in terms of a uncertain decisions in whole process design. plurality of prioritisation vectors. Until now, The new method utilises simulations based the multiplicity of prioritisation vectors upon triangular distributions to create a plot supplied by the Spanning Trees approach which visualises the preferences and have been amalgamated into a single priority overlapping uncertainties of decision vector, losing the information about the alternatives. It facilitates decision-makers to 74 visualise the not-so-obvious uncertainties of so, given a pairwise comparison matrix A, we decision alternatives. SURE has been devise some optimization problems to find its compared to other known methods such as closest transitive approximation X. The AHP and ELECTRE and has been shown to distance between the matrices A and X can perform better than these existing methods. then be used as an estimation of the degree The SURE method has been implemented in of intransitivity of the matrix A. In other R as a part of the MCDA package. A words, we consider the distance between A demonstration will be shown of how to apply and the set of all transitive pairwise SURE (using the MCDA package in R) in a comparison matrices as an index of real-world case study for a large intransitivity. This new approach is justified pharmaceutical company. by means of some numerical examples and the optimization problems used to find the 4. Distance-based Measures of closest transitive approximation to a pairwise Incoherence for Pairwise comparison matrix A can be formulated as Comparisons mixed-integer linear programs. In our study, we consider the usual condition of transitivity *Matteo Brunelli, University of Trento, but also restricted max-max transitivity and Italy, [email protected] weak consistency. Finally, we test and Bice Cavallo, University of Naples validate our approach from a numerical point "Federico II", Italy, [email protected] of view and we use it to calculate the intransitivity of some pairwise comparison In our research we deal with pairwise matrices which have already been analyzed comparison matrices. They have been mainly in the literature. With our approach we hope employed in the Analytic Hierarchy Process to offer a more natural methodology to (AHP), but their use is important in many quantify the degree of intransitivity of other MCDM methods too, as for instance cardinal preference relations and pairwise Multi-Attribute Value Theory (MAVT). Once comparison matrices. the preferences of a decision maker have been collected into a pairwise comparison TUE-1-E matrix, testing their coherence is a crucial step. The most widely used and tested Invited Session: Multiobjective condition of coherence is the condition of Optimization Applications consistency. Many inconsistency indices Tuesday 9:00-10:40 - Room: Hagia Sophia have been proposed in the literature to estimate the rationality of preference. Their Chair: Banu Lokman final goal is to isolate preferences which are too irrational and possibly ask the decision 1. Forming a Collaborative and maker to rethink them. In real-world Productive Team of Experts situations, it is almost impossible to reach *Mehdi Kargar, Ryerson University, full consistency. Thus, conditions weaker than consistency have also been formulated. Canada, [email protected] Among them, one can consider various The success of any project is directly related transitivity conditions. So far, the violation of to the team that perform and execute the these conditions has been measured with project. Forming the most effective team of simple counts of how many times a given experts is a challenging task for many condition is violated by a set of preferences. institutes and organizations. When forming a It is reasonable to say that, when we deal team, collaboration and productivity are with cardinal preferences, this remains a among the most important factors that simplistic approach, since it calculates how guarantee the success of the team. Given a many times a condition is violated, but not social network of experts, we tackle the the extent to which it is violated. That is, a problem of forming a team of experts that simple count only employs ordinal covers a set of required skills. An expert information discarding the precious possesses a set of skills. These skills are information contained in the cardinal obtained by the expert through training or preferences. In this research we propose a past experience. Furthermore, based on past cardinal approach to the measurement of collaboration among experts in various the violation of transitivity conditions. To do projects, a graph among experts can be built.

75

If two experts have past collaboration (e.g., project. The expertise of an expert is work on the same project), there will be an extracted from the titles of the projects she edge between them in the underling graph. participated in. The results suggest that the This graph can be weighted. If two experts proposed evolutionary algorithm is more collaborated in more than one project in the stable in terms of run-time than the greedy past, the edge weight between them will be algorithm when the frequency of the required smaller than two experts that only skills changes. Furthermore, the results of collaborated in one project. Given a project our proposed method achieve high accuracy that needs a set of required skills (i.e., when compared to the exact algorithm expertise), the goal is to form a team of (exhaustive search). experts that together cover the required skills and have past collaboration. Past 2. Incorporating Robustness in collaboration can be measured by the weight Interactive Multi-Response Surface of the sub-graph that connects experts Optimization of Product Design together. In this case, the weight of the sub- Parameters graph is called the communication cost among experts. The smaller the *Melis Özateş Gürbüz, Middle East communication cost, the better the team. Technical University, Turkey, The problem of finding a team of experts [email protected] while minimizing the communication cost Gulser Koksal, Middle East Technical has been studied in the past. However, each University, Turkey, [email protected] expert has a level of productivity. Productivity Murat Köksalan, Middle East Technical of each expert can be obtained by the University, Turkey, number of projects that have been completed by her or the quality of the [email protected] projects that she has delivered in the past. In Several factors affecting quality of a product this case, each expert will be assigned with a have to be considered to determine settings productivity score. It is desired that a team of of design parameters that yield high quality experts for performing a given project has with minimal variation (design robustness). high productivity. Optimizing both The quality of a product need to be defined productivity and communication cost turns by multiple responses where optimal design the problem into a bi-objective optimization solution for a response is typically not problem. This problem has not studied in the optimal for another due to the conflicting past. Since optimizing each of these natures of responses. These problems can objectives are NP-hard, we propose a be referred to as Multi Response surface knowledge-based evolutionary algorithm to Optimization (MRO) problems. The find a team of experts that optimize both relationships between responses (or these objectives. We also propose a greedy performance measures) of interest and algorithm solve this bi-objective optimization relevant design parameters are modeled as problem. To demonstrate the effectiveness empirical relationship functions. Finding and efficiency of our approach, we compare meaningful solutions requires considering it with the results of the exact algorithm that such multiple response functions is achieved by the exhaustive search simultaneously as objectives or constraints algorithm. The experiments are performed in optimization models. Many of the existing on the real DBLP and GitHub datasets. approaches in the literature optimize a DBLP is a publication dataset that contains specific function of the responses, implicitly information about authors of scientific assuming that all DM’s have the same papers. The experts are researchers that preferences. Others assume the preferences publish papers. If two authors publish of the decision maker (DM) can be together, they will be connected in the represented by a specific function and its underlying graph. Productivity can be parameters are known exactly at the outset. measured by the number of papers that are These are strong and unjustified published by each author. GitHub is a hosting assumptions in practice. There are a few service for IT projects. Nodes of the GitHub studies that employ interactive multi graph are developers that participate in objective optimization (MOO). Although these different projects. Two experts are connected approaches incorporate the DM’s to each other if they participate in the same preferences into the solution process, they

76 typically ignore model uncertainty. The respective target values. The DA identifies uncertainty inherent in the structure and/or solutions that may be attractive and presents parameters of the response surface models several of them to the DM. The procedure is important and solutions that do not take continues until the DM is satisfied. We this into account may result in designs that demonstrate our approach on the Polymer are unstable and hence undesirable to the experiment, a two-response problem widely DM. Another common assumption in the used in the literature. literature is that responses have constant variances at different settings of the 3. A Multi-Objective Approach to controllable design parameters. This Cluster Ensemble Selection Problem assumption may not be realistic for MRO *Dilay Aktaş, Middle East Technical problems and the solutions obtained based on these assumptions may have properties University, Turkey, [email protected] that could be substantially different from the Banu Lokman, Middle East Technical estimated properties. We develop an University, Turkey, [email protected] interactive approach for the two-response Tülin İnkaya, Uludag University, Turkey, product and process design optimization [email protected] problem, considering DM’s preferences under model uncertainty associated with the Clustering is an unsupervised learning parameters of response surface models and method that partitions a dataset into groups. heterogeneity of response variances. We use The aim is to assign similar points to the several controllable properties of the same cluster and dissimilar points to responses, such as the distance of the different clusters with respect to some notion estimated response means from their of similarity. It is applicable to a wide range respective target values and the estimated of areas such as recommender systems, response variances, to control the anomaly detection, market research, and performance of the solutions in our models. customer segmentation. With the advances In order to provide relevant information on in the computational power, a diverse set of the consequences of solutions to the DM, we clustering solutions can be obtained from a produce visual aids on performance dataset using different clustering algorithms, measures such as joint response confidence different parameter settings and different and prediction regions of selected solutions features. Clustering ensemble has emerged together with lower and upper specification as a powerful tool for combining the limits and target values of the responses. strengths of these multiple clustering The performance measures and visual aids solutions and generating a consensus facilitate the communication with the DM in solution. It improves the quality of clustering extracting preference information. Although in terms of accuracy, robustness and some general knowledge on the effects of stability. In this study, we address the cluster changing the controllable properties on the ensemble selection problem, and propose a performance measures exist, the exact multi-objective approach to generate a magnitudes of those effects are not known consensus clustering solution. The proposed since they are determined by the complex algorithm selects a representative subset of problem characteristics. It is not possible to clustering solutions, and produces a know the exact effects of sacrifices on some consensus clustering solution by combining performance measures with certainty. these representatives. Different from the Therefore, we involve the design analyst (DA) existing approaches, we design the in the search process to help the DM representative selection approach based on converge to his/her preferred solutions three criteria: quality, diversity and size of the quickly. At each iteration, the DA converts representative set. Before the representative DM’s verbal preferences into mathematical selection, we apply a preprocessing expressions that have the potential to lead procedure to analyze the characteristics of the search towards regions containing the clustering solutions in the full ensemble solutions that are desirable to the DM. The and eliminate the ones that may mislead the DA systematically searches the relevant consensus function. We test the solution space utilizing Pareto frontiers performance of the proposed approach on considering only the distances of the benchmark datasets. The results show that estimated response means from their the proposed approach works well, and it

77 outperforms the existing methods. TUE-1-F Furthermore, we observe that regardless of the ensemble generation mechanism and Contributed Session: Understanding consensus function, the preprocessing Consumers and Markets procedure increases the accuracy of the final Tuesday 9:00-10:40 - Room: Basilica Cistern consensus clustering with respect to true cluster labels. Keywords: Cluster ensembles, Chair: Fusun Ulengin consensus clustering, multi-objective clustering 1. A Comparative Analysis of Entrepreneurial Success of 4. A Web-based Decision Support Advanced and Emerging Economies: System for Multi-Objective Mixed A SEM Approach Integer Programs Seda Yanık, Istanbul Technical *Banu Lokman, University of University, Turkey, [email protected] Portsmouth, United Kingdom, *Nihat Can Sınayiş, Istanbul Technical [email protected] University, Turkey, Gokhan Ceyhan, Middle East Technical [email protected] University, Turkey, [email protected] Today entrepreneurship and start-up Murat Köksalan, Middle East Technical companies are the most important source of innovation and economic growth. Thus, the University, Turkey, ecosystems to foster entrepreneurship is [email protected] emphasized more and more in almost all Multi-objective mixed integer programs economies. Efforts for increasing the (MOMIPs) are widely used to represent and entrepreneurship can only be successful model the decision problems in many when the root causes and their relationship organizations. It is typically hard to with the entrepreneurial success are characterize the nondominated set of understood. This paper is an investigation MOMIPs especially in complex systems. In of the main factors of creating successful this study, we develop a web-based decision new businesses and a comparison between support system that provides users with the two main country groups of Europe. tools to solve multi-objective integer and Countries with advanced economies and, mixed integer programming problems. Our emerging and developing economies will be first algorithm in the web-based solution set side by side to understand whether the platform is designed to generate all pre-indicated factors can be generalizable nondominated points for a given multi- for these groups. In this study, we objective integer program. In order to reduce hypothesize the definition of entrepreneurial the computational burden, we also develop success as a highly relatable indication to algorithms to generate representative main five factors. These are the nondominated points for any given MOMIP. infrastructure, economic situation, These algorithms are designed to generate a educational structure, culture and market subset of nondominated points that dynamics of a country and is investigated to represent the whole nondominated frontier form the basis of success in new ventures. with a desired quality level. We also develop We propose a model with the visualization tools that provide users with the interrelationships of these factors. We opportunity to analyze the distribution of the estimate the model using structural equation nondominated points over the objective modeling and test our hypotheses. Historical function space and to visualize the tradeoffs data is used to test whether the chosen between the objective functions. We also variables of five main factors create a maintain a digital library that provides a significant relationship with entrepreneurial collection of test data sets and make their success. According to the results of the inputs and outputs available to researchers. model estimation, variables with no effect are taken out and a final result is reached. Data that is used in this research is obtained from Global Entrepreneurship Monitor’s (GEM) two databases: National Expert Survey

78

(NES) and Adult Population Survey (APS). the minimization of the discomfort NES provides the national perspective on associated with changes regarding most new businesses and Adult Population Survey preferred settings or operation time slots. (APS) supplies the social viewpoint of The model considers different types of loads: individuals towards entrepreneurship. We - shiftable loads, i.e., loads for which the estimate the model separately for the data of operation cycle cannot be interrupted such the country cluster of advanced economies as dishwashers, tumble dryers or cloth and for the data of the country cluster of washers; - interruptible loads, i.e., loads emerging economies. Then, we compare the whose operation cycle can be interrupted, relationships of the factors that lead to given that a certain amount of energy is entrepreneur success for these two different supplied before a specified instant, such as clusters of countries. Our findings can be electric water heaters or electric vehicles; - used by policy makers to identify the thermostatic loads, i.e., appliances potential improvement areas for increasing controlled by a thermostat device switched the entrepreneurship and its success. by the indoor temperature, which is Moreover, the results can also provide a determined by a thermal model of the space benchmarking basis for emerging countries. being conditioned, such as air conditioners. Constraints are related to operational issues, 2. A Bi-Objective Cost-Comfort MILP such as avoiding load diagram peaks, and Model of Demand Response to quality of service / comfort requirements, Dynamic Tariffs such as respecting time slots for appliance operation or bounds for indoor comfort *Carlos Henggeler Antunes, University temperature. Scalarizations of the model of Coimbra, Portugal, [email protected] instantiated with realistic data can be solved Maria João Alves, University of Coimbra, with a commercial solver for 15-minute Portugal, [email protected] discretization of the planning period, enabling to characterize the nondominated Dynamic time-of-use (ToU) tariffs are front. For finer time discretization, the model expected to become a common pricing cannot be solved exactly due to its scheme in smart grids, by conveying price combinatorial nature and, therefore, signals reflecting generation costs, e.g., customized meta-heuristic approaches are involving the availability of renewable energy, necessary to obtain good quality solutions in and grid conditions, e.g., congestion. an acceptable computational time, which is Therefore, ToU tariffs offer consumers compatible with (near) real-time incentives to adopt energy usage patterns implementation as an automatic home different from habitual behaviors. In this energy management system. In a smart grid setting, ToU tariffs can bring benefits for: - setting, these systems using a low-cost grid operators by contributing to microcontroller are expected to be deployed accommodate larger shares of intermittent in the premises of consumers to control end- renewable generation and alleviate use loads by responding to different grid congestion in networks adopting a “load requests while considering end-users’ follows generation” paradigm; - electricity preferences. Computational results are retailers by enabling to manage wholesale presented displaying the characteristics of buying and retail selling prices; - consumers solutions on the nondominated front, by engaging in demand response actions to emphasizing the trade-offs between the reduce the energy bill with jeopardizing economic and comfort dimensions for comfort. Consumers react to dynamic tariffs different consumer profiles. The detailed by resetting thermostat setpoints or operation of appliances is presented as well rescheduling appliance operation to profit as comfort indicators such as indoor from lower priced periods, thus balancing the temperature along a one-day planning cost and comfort evaluation dimensions period. It is shown that shiftable and associated with the provision of energy interruptible loads are easily dealt with by the services. This work presents a bi-objective solver, but the control of the air conditioning mixed integer linear programming model system imposes a significant computational (MILP) to optimize demand response in face effort. Also, the MIP gap increases with the of dynamic tariffs, considering the weight assigned to the indoor temperature minimization of energy and power costs and

79 comfort objective, which may impair real- variables at the columns; and average RFM time applications. values of the customers in each cluster at the entries. In order to compute the CLV scores 3. Assessing CLV Scores of the of the clusters; obtaining a global score by Customer Segments Through a adding weighted contributions from each Weighted RFM Decision Model variable is frequently preferred in the literature. Hence, at the last step, we used *Berna Unver, Sabancı University, Simple Additive Weighting (SAW) method for Turkey, [email protected] this purpose. As a result, we came up with Füsun Ülengin, Sabancı University, CLV scores of each cluster. Those scores are Turkey, [email protected] used to differentiate the corresponding Ilker Topcu, Istanbul Technical clusters and to suggest managerial insights University, Turkey, [email protected] for each of them. This study is related with a significant stage 4. Analysis of Marketing Strategic of a research project, which aims to provide Planning in Furniture Industry by a DSS that analyzes the big data related to Applying ANP, Case of the Study: Iran retail consumers in order to describe and predict their purchase behaviors and Furniture Industry propose managerial insights for the retail *Majid Azizi, University of Tehran, Iran, sector managers. These insights are [email protected] expected to improve the business processes of the companies and their performances, This study was carried out due to the severe create new markets and opportunities as dearth of research on strategic planning in well as ensure sustainable competitive the country's furniture market. Accordingly, advantages. In the context of that project, a this research represents a decision-making case study was conducted on datasets model developed to select the best solution consisting the transactions of individual and for strategic planning in the industry. Four business members of Sam’s Club, a division possibilities can be considered as potential of Walmart Stores, Inc. In this study, after solutions: Entrance to the foreign market clustering individual and business members (S1), Increase portfolio (S2), Emphasis on of Sam’s Club into customer segments scientific management of mixing elements of utilizing descriptive and predictive analytics, marketing (S3), Generate research and we aimed to estimate the customer lifetime development units (S4). The Analytic value (CLV) of the clusters based on the Network Process and the Super Decision purchasing behaviors of the members in software were used to synthesize and each cluster. Companies can use CLV as a analyze the model. It was found that all metric for determining which customer calculated decisions were influenced by segments are more profitable and loyal as strategic criteria. A value-weighted well as developing efficient and appropriate competency model was calculated in the first marketing and sales strategies for each stage with the influence of strategic criteria segment with respect to financial and on the competency model. Hierarchical operational perspectives. To compute CLV designs decisions were made for each of the score for each cluster, we used a weighted competencies and their subsets (144 sub- RFM (recency, frequency, monetary) decision criteria and 22 middle indices). Paired model. In the first step, to assess the relative comparison matrices associated with the weights of RFM variables, we assessed degree of importance of each of the judgements of the experts by asking pairwise competencies were achieved in the second comparison questions, formed a pairwise stage. In the final stage, subsets of comparison matrix using those judgements, competencies’ weights and their sub-options and used the values of the eigenvector as were identified with the combination of the relative weights extracted from that matrix. competencies’ networks. The results showed In the second step, we constructed two after collecting questionnaires, evaluating decision matrices, one for individual and correcting the incompatibility rate of customers and the other for business comparative matrices, with respect to customers. Matrices have clusters at the competencies of benefits, opportunities, and rows; recency, frequency, and monetary costs, third solution (S3) has the highest

80 priority, while regarding the competency of 1. Entrepreneurship Education: risks, first solution (S1) has the highest Towards The Validity of an Impactful priority. The final results showed that the third solution (S3) is the best alternative and Effective Framework for among the alternatives proposed in the Business Schools marketing strategy of the home furniture *Olga Porro, ESADE - URL, Spain, industry. [email protected] TUE-2-P Monica Sanchez, UPC BarcelonaTech, Spain, [email protected] Plenary Session 2 Núria Agell, ESADE-URL, Spain, Tuesday 11:10-12:10 - Room: Mustafa [email protected] Kemal Amfisi Marcel Planellas, ESADE - URL, Spain, Chair: Serpil Sayın [email protected] Multi-Objective Combinatorial The nature and characteristics of entrepreneurial competencies have been Optimization – Beyond the Bi- widely investigated and applied to many Objective Case fields, such as entrepreneurship research *Kathrin Klamroth, University of (Man et al 2002), competency and competence theory or business performance Wuppertal, Germany, and growth (Mitchelmore and Rowley 2010). [email protected] The most common entrepreneurial teaching Multi-objective combinatorial optimization model of a business school is a combination (MOCO) is a quickly growing field that is of entrepreneurial theoretical courses, highly relevant for a multitude of application engagement in practical oriented activities, areas and at the same time highly offering of innovative spaces or mentoring challenging due to its inherent complexity. support, among other features (Sirelkhatim Typical examples of MOCO problems include and Gangi 2015). Based on the existing multi-objective knapsack and assignment literature, there is no blueprint on a valid problems, the multi-objective TSP, and structure or hierarchy of these elements with network problems like multi-objective respect to their relevant importance minimum spanning tree, shortest path, and regarding the impact on the future minimum cost flow problems. Starting from entrepreneurial path of students (Lorz et al a review of recent developments in the field 2013). The objective of this paper is to of MOCO, we will analyze the main drivers of investigate, in a business school context, the the complexity of MOCO problems and influential variables and sub-variables of an identify cases where MOCO problems are education model that effectively enhance actually easy. We discuss general solution and promote entrepreneurial competences concepts such as generic scalarization among students. The list of variables based algorithms, branch and bound presented below is the result of a process of methods, and upper and lower bound sets. information gathering, compiled through 12 This leads to the question of concisely deep interviews to academic experts in the describing the search region and efficiently field. Variable 1: Curricula (Sub-Variables: updating intersections of polyhedral cones in Specific knowledge on Entrepreneurship; the higher dimensions. use of business cases; the use of inspirational sources (bring testimonials to TUE-3-A class); Business Plan as a final project). Variable 2: Institutional support (Sub- Contributed Session: Multiple Criteria variables: Innovative physical spaces and Ranking or Sorting related resources; Clubs or student Tuesday 13:50-15:30 - Room: Galata Tower associations; Alumni networking; Institutional financial support). Variable 3: Chair: Dalel Amami Real (outside) immersions (Sub-variables: Visits; Study tours o Immersion trips). Variable 4: Participation in challenges (Sub- variables: Challenges and competitions

81 proposed by real companies; Simulated advantage, natural adaptation strategies challenges and competitions). Variable 5: push companies to obtain more sustainable Participation in practical activities (Sub- policies facing climate change. Low CO2 variables: Service Learning; Internships in emissions, less harm to the environment, start-ups). These variables are ranked with etc. become the new strategical challenges respect to their impact and effectiveness to to overcome. Having a high tech delivery enhance and promote entrepreneurial service with maximum information share competences. Data is obtained from the with the customer in addition to the Entrepreneurship Institute of ESADE minimum harm on nature is the most Business School in Barcelona. The groups of important trend in last mile delivery. To be decision makers considered are: First, able to reach that, different strategies have Alumni Entrepreneurs who have set up their been suggested. Electric vehicles (EVs) are own start-up. Second, Alumni intra- one of the most powerful solutions. EVs are entrepreneurs or intra-innovators, within a the energy efficient choice for inner-city bigger company. And, finally, “Life” Alumni traffic. Transformation of conventionally who started an entrepreneurial project fueled vehicles to EVs aids to reduce both related to civil society, sports or personal life, emissions and noise for the urban beyond their professional scope. All congestion. On the other hand, EVs are not respondents will have the possibility to highly preferable for the delivery companies assign hesitant values if they are not sure due to insufficient infrastructure for charging about their judgment. network, high investment costs and lack of information. However, technology continues 2. Electric Fleet Selection for Last to grow, and smart cities are not just future Mile Delivery dreams. Strong charging networks have been developed in various cities, and they *Deniz Uzturk, Galatasaray University, will be more reachable shortly. Therefore, Turkey, [email protected] this paper focuses on the electric fleet Gülçin Büyüközkan, Galatasaray selection problem for a delivery company. It University, Turkey, is undeniable that EVs are the future vehicles [email protected] and many European countries prepare to Fahri Negüs, Galatasaray University, stop fueled vehicle sale by 2025. Getting Turkey, [email protected] ready for the future with competitive Yaman Oztek, Galatasaray University, strategies are very critical for delivery Turkey, [email protected] companies. In this study, the main focus is to determine the evaluation criteria for the Last mile delivery is the ultimate chain in electric fleet selection and approach this urban logistics. It refers the micro-level selection procedure as a multi-criteria logistics activities from the delivery center to decision making (MCDM) process. the end-user. It has a very complicated Evaluation based on distance from the relationship with urban transportation, and it average solution (EDAS), which is firstly needs to be evaluated together with city introduced by Ghorobaee et al. in 2015, tool transportation to be efficient and has been provided for the selection problem. sustainable. E-commerce, which is the It has been used with its 2-Tuple extension in powerful catalyst for last mile delivery, is order to create a convenient environment to overgrowing day by day. This growth use linguistic variable during evaluation. uncovers the need for more effective and on- EDAS method has been preferred thanks to time deliveries for the companies. Delivery its efficient and straightforward companies try to distribute the good to the computational steps while handling multiple right place at the right time. The one that criteria. The most suitable alternative is delivers it faster takes a competitive selected according to the average solution advantage; therefore having a rapid and that is determined by arithmetic mean. By efficient last mile delivery is a very crucial extending the classical EDAS method with a strategy for the company. Also, the more 2-Tuple linguistic model, it becomes more technology is integrated with the delivery suitable for group decision-making (GDM). process; the more customers want to be Usually, fuzzy extensions of EDAS also flexible and connected to the on-time provide convenience for GDM; but when information. Besides the competitive different levels of knowledge are included in

82 the group different linguistic levels are governmental dimensions. The citizen’s needed. Hence, the 2-Tuple linguistic model perception of quality of life within the city is has the power and simplicity to aggregate an important information, which serves as a different granulated information raised from basis for improving certain urban various decision makers (DMs). To conclude, performances, which citizens perceive as this study proposed an MCDM approach for insufficient in their cities. For the analysis of electric fleet selection subject. 2-Tuple various attributes of cities’ urban integrated EDAS has been used to select the performance, a multi-criteria model is most suitable alternative. Also, Best Worst developed, integrating the entropy for Method (BWM) has been suggested with its determining the criteria’s relative weights 2-Tuple extension to weigh the criteria for the and VIKOR (VIšeKriterijumska Optimizacija I selection problem. Finally, a case study is Kompromisno Rešenje, Eng. Multi-Criteria conducted to see the plausibility of the Optimization and Compromise Solution) as suggested MCDM methodology. The results ranking method. Respondent’s subjective and the discussions are provided in the preferences were measured by 4-point Likert study. scale and for the purpose of the analysis the quantification of responses was made. The 3. Assessing Smartness and Urban multi-criteria model provided in the paper Development of the European Cities: includes 28 different criteria with An Integrated Approach of Entropy approximately equal significance, grouped and VIKOR into six categories and from 73 up to 111 European cities per year depending of the Jelena Stanković, University of Niš, sample size in each of the research cycles. Serbia, The results of the ranking are visualised in [email protected] order to determine urban areas in Europe *Žarko Popović, University of Niš, where the inhabitants perceive is the highest Serbia, [email protected] overall level of smartness and sustainable Ivana Marjanović, Innovation Center urban development. The main finding of the paper implies that among the best ranked University of Niš, Serbia, cities are mostly cities of Western Europe, [email protected] mainly United Kingdom. It should also be Contemporary flows put urban areas at the noted that apart from Vienna, Luxembourg, forefront as the main bearers of economic Copenhagen and Stockholm, the other growth and prosperity. The accelerated analysed capitals are not in the top 10 in expansion of urban areas in last decades each of the observed years. poses the issue of their sustainability and 4. Application of Multi-Criteria growing infrastructural, ecological and social problems. The concept of smart cities is one Decision Making for Patients Priority of the potential responses, strived at and Assignment to Operating Room increasing the competitiveness of local communities and urban areas, through the *Dalel Amami, University of Sfax, application of innovative technological Tunisia, [email protected] solutions in order to improve the quality of Fatma Dammak, University of Sfax, life for its citizens providing better public Tunisia, [email protected] services and a healthier environment. The Successful assignment and scheduling paper is aimed to perform ranking of the patients to the operating room are still European cities according to their smart and problematic matters. To solve these issues, urban development indicators. The research several methods were proposed such as results are determined based on the data Mathematical programming, heuristic, collected through four cycles of the metaheuristic. Multicriteria decision-making Eurostat’s Urban Audit Perception Survey (MCDM) tools could also be useful in finding conducted in the period 2006-2015. The solutions. MCDM proposes a set of data refer to perceived urban performances techniques applied to certain criteria to help of cities by inhabitants living in them, the decision-makers (DMs) identify, compare including different aspects of sustainability and estimate alternatives. The best and development of urban areas such as alternative should answer the constraints of economic, social, environmental and

83 the DMs and their preferences and priorities 1. A Multi-Criteria Decision Making The patient assignment problem can be Approach to Identify the Priority of assimilated as a multicriteria decision- making problem where many criteria should Criteria Affect Sustainability of Eco- be considered in decision-making. This paper Villages in Turkey proposes a novel approach that is the *Gökçen Bayram, Marmara University, application of MCDM for assigning patients Turkey, [email protected] to the operating rooms. Therefore, we address the problem of operating room Recent economic, ecologic and social planning and scheduling by using a multi- challenges lead to small and self-sufficient criteria approach. The production goal is to habitats, called eco-villages to deal with assign a list of patients on a limited number challenges and reduce the negative impact of operating room subject to a set of criteria. of pollution. We now see an increasing We propose an approach based on the use interest in the adoption of establishing and of C-TOPSIS with different weight methods. living in an eco-village all around the world. We will respectively use Entropy Method During 2000s there has been several (EM), the standard deviation (SD), and attempts to establish eco-villages but now Preference selection index (PSI) to determine there are a few eco-village initiatives in the important scores of evaluation criteria. Turkey. The term of wellbeing, which can be The objective of this procedure is to minimize defined as all aspects involves the the makespan of the operating room, taking contribution of an individual’s and the into account the priority of the patient. Recall community’s synergy and togetherness that the makespan denoted Cmax is the positively, is the key fact to provide maximum total processing time on any sustainability of eco-villages. In this paper, a machine. The processing time is the multi-criteria decision making model is operation duration; the machines are the proposed for identification of sorting operating rooms. The obtained results show according to priority of criteria that affect that the best value of Cmax is obtained by sustainability of eco-villages in Turkey. The the PSI method. This result can be explained study is structured on three major steps; by the ability of the PSI to deal with different criteria identification, questionnaire survey criteria by using the fuzzy logic to study and MCDM method implementation. In homogenize the different criteria functions this study Analytic Hierarchy Process (AHP) is Our proposed technique has been applied to used as the methodological tool to evaluate generate multiple schedules taking into and to identify the priority of the criteria. account the different considered criteria. Literature review is utilized to determine This evaluation method is used to select the criteria and the study is based on 5 criteria best solution satisfying the preference of the and 43 sub-criteria. In the extensive decision-maker based not only on the questionnaire survey the criteria and sub importance of each criterion but also on the criteria are graded by using 1-9 Saaty scale interactions between the criteria. We note in order to compare each criterion. the shortage of research considering the Participants are selected from the resolution of the scheduling problem by experienced with eco-villages as inhabitant using TOPSIS technique, further work needs in an eco-village and experts in several to focus on solving the proposed problem ecology associations. The proposed with other MCDM approaches in order to evaluation criteria provide a reference for compare them with the presented approach. eco-village inhabitants and founders to provide sustainable eco-villages. TUE-3-B 2. The Analytic Network Process as a Contributed Session: AHP/ANP Planning Tool for Sustainable Applications in Urban and Regional Development Development Tuesday 13:50-15:30 - Room: Maiden's *Hannia Gonzalez-Urango, INGENIO Tower (CSIC-UPV), Spain, [email protected] Chair: Sercan Cakir

84

Monica Garcia-Melon, Universitat processes. The achievement of an optimal Politecnica de Valencia, Spain, solution for all of the stakeholders becomes [email protected] difficult when the intervention of different agents, objectives and factors is considered, 1. Introduction This study conducts a which involves the interaction of complex systematic literature review on the Analytic elements in complex context. 3. Hypotheses Network Process's support for decision of the work This work is based on the next making processes for sustainable hypothesis: ANP is a useful tool to support development. It focusses on the usefulness planning processes for sustainable of the methodology and draws some development. 4. Research conclusions on its potential and limitations. Design/Methodology There are five steps The research is frameworked in sustainable that we propose to achieve the systematic development planning. Several authors literature review (Denyer & Tranfield 2009): introduced the use of MCDM techniques for 1. Question formulation 2. Locating studies: sustainability issues (Ginevičius and Identification 3. Study selection and Podvezko 2009). Many of them focus on the evaluation: Screening 4. Analysis and use of the Analytic Hierarchy Process (AHP), synthesis 5. Reporting the results which has been stood out as the most often used (Dos Santos et al. 2019) due to its ease References Ginevičius, R., and Podvezko, V. of use for preferential information elicitation (2009). “Evaluating the changes in economic from expert. However, AHP does not allow and social development of Lithuanian considering the interdependences among counties by multiple criteria methods.” Baltic criteria. For this reason, some authors Journal on Sustainability, 15(3), 418–436. recommend the use of the Analytic Network Dos Santos, P. H., Neves, S. M., Sant’Anna, Process (ANP) because it allows better D. O., Oliveira, C. H. de, and Carvalho, H. D. representing the complex interactions (2019). “The analytic hierarchy process among the different components of a real supporting decision making for sustainable system (Wu and Cui 2016) as is the case in development: An overview of applications.” the field of sustainability. As a result, several Journal of Cleaner Production, Elsevier Ltd, authors have introduced the use of ANP 212, 119–138. Wu, X., and Cui, P. (2016). “A during the last decade. Regarding Study of the Time–Space Evolution sustainability, many works in different areas Characteristics of Urban–Rural Integration have been proposed, e.g. engineering, Development in a Mountainous Area Based environmental sciences ecology, science on ESDA-GIS: The Case of the Qinling-Daba technology, business economics, operations Mountains in China.” Sustainability, 8(11), research management science, computer 1085. science and energy fuels. This work gathers, 3. A Combined Simulation- maps, analyzes, and summarizes this academic literature by reviewing Optimization and AHP Approach to manuscripts in which ANP is applied for Improve the Quality of Care in sustainability and which are indexed in the Healthcare Centers Web of Science and Scopus databases. The results, analysis, and discussion of the *Mohammad Dehghani, Northeastern screened manuscripts show (a) a descriptive University, United States, analysis, (b) bibliometric data, (c) a citation [email protected] analysis, (d) an analysis of the most relevant Nihan Kabadayi, Business School of literature and (e) an analysis of the most Istanbul University, Turkey, recent literature. Results also allow [email protected] comparing the findings of this research with the work proposed by Dos Santos et al. Quality of care is crucial for patients' (2019) regarding AHP. 2. Sustainable satisfaction and safety in healthcare centers. development in planning process The The majority of hospitals attempt to definition of sustainability implies a three- implement facility-wide improvements to pillar concept: environmental, economic, and ensure high-quality care delivery. This study social. However, some recent publications in aims to propose a combined Simulation- the field of sustainable planning also stress Optimization and MCDM approach to the importance of including participatory accurately asses the impact of quality

85 improvement initiatives on different facets of contribution, as we consider multiple criteria healthcare systems. In this framework, first, decision making (MCDM) methods as one the importance (weights) of the different potential solution in order to tackle the healthcare criteria is determined by health ambiguity in the selection of suitable urban providers’ using an AHP approach. Then, the green spaces. The study preliminarily weights provided by AHP are applied in a focuses on comparing two prominent MCDM simulation-optimization environment to methods which are the Analytic Hierarchy determine the most efficient action with the Process (AHP) and Multi-Attribute Utility most desirable quality of care. Simulation Theory (MAUT). We will discuss the suitability provides a platform to examine the of both MCDM approaches in terms of effectiveness of different improvement acquisition and modelling of the user efforts and calculate their impact on the preferences and the computing of the system performance measures. weights. The a priori knowledge of user preferences for different activities will be 4. An MCDM Approach for Selecting collected through questionnaires. Both Suitable Urban Green Spaces Based MCDM methods are then implemented for a on a Desired Activity specific use case scenario (jogging activity). In this scenario, each evaluation criterion, for *Sercan Cakir, Leibniz Institute of instance, a) the size, b) the path network Ecological Urban and Regional length, and c) the green volume of each Development, Germany, [email protected] urban green space related to activity Robert Hecht, Leibniz Institute of “jogging” are associated with weights which Ecological Urban and Regional are estimated based on the user’s personal Development, Germany, preferences. Afterwards, the scores of urban [email protected] green spaces are calculated by aggregating Kerstin Krellenberg, Leibniz Institute of the multiplication of indicators on the normalised values and estimated weights. Ecological Urban and Regional Here we aim at providing a novel approach Development, Germany, for applying MCDM to urban green space [email protected] activity patterns. The results obtained from Publicly accessible urban green spaces play both techniques will be presented and an essential role to ensure a high quality of discussed in terms of the strengths and life in cities by providing citizens with spaces limits. for recreation and social interaction. TUE-3-C Particularly in cities with high growth rates and the often associated densification, it is Invited Session: MCDM/A Models with especially important for citizens to know Flexible and Interactive Preference where urban green spaces are, how they are Modeling equipped and how they can be accessed. With this information, the benefits of Tuesday 13:50-15:30 - Room: Dolmabahçe ecosystem services, e.g. cultural services Palace such as recreational experiences, but also Chairs: Suzana Daher, Rodrigo J P Ferreira regulating services such as the cooling effect through vegetation, which citizens receive 1. The MCDA as a Way to Avoid from urban green spaces, can be properly Political Polarization in Social exploited. In this context, a service Elections infrastructure and a web application for cities are being developed within the framework of *Pablo Manyoma, Universidad del Valle, an ongoing research project. One entry point Colombia, of the app is to provide users with [email protected] information on suitable urban green spaces within a city based on a desired activity such Political polarization is a widely known as jogging, walking or reading. This requires phenomenon that has always attracted the the development of a decision support attention of researchers throughout history. system for the selection of urban green It has been believed that dialogue through spaces for specific activities. How this can be different political lines is a key prerequisite developed is part of this conference for maintaining democratic citizenship. One 86 could even say that the debate is system of choice is proposed that allows irreplaceable to form rational opinions that ordering from highest to lowest the go beyond the limits of one's own subjectivity preference that citizens have for each to incorporate the points of view of others. In candidate in each aspect declared as addition to its importance in everyday life, important. This is done through the Borda political polarization has been the subject of system of social choice. This methodology debate and academic exploration due to its would reduce the polarization of candidates potentially negative consequences in and would force the voter to be more aware democratic societies. These consequences of government programs proposed by the range from simple discomfort among people candidates themselves. to the use of this polarization as an element of retrospective economic evaluation, 2. Understanding a Decision Policy: whether in recessions or recoveries in a Using Fitradeoff as a Mirror for a particular country. The cases in which the Decision Maker supporters see the opposite party more negatively than theirs have increased in *Leonardo Antonio Pessôa, CASNAV, recent years in the Americas. The cases of Brazil, [email protected] the United States, Mexico, Costa Rica, Rodrigo Ferreira, CDSID/UFPE, Brazil, Colombia and Brazil show that, many times, [email protected] polarization is a problem of preferences. In Carla Lage, CASNAV, Brazil, this way, the extremes take on more [email protected] importance and the midpoints are rarely Adiel Teixeira de Almeida, Federal seen. When this political scenario is University of Pernambuco, Brazil, configured, a complex series of phenomena begins to occur. Normally in these situations, [email protected] the poles dominate the social conversation The paper provides an evaluation of a in all areas, they make their actions acquire decision policy, made by consensus among a more weight in public opinion, concentrating group of high-level decision makers of the all available resources, increasingly, around Brazilian Navy. There is not an unique set of these two poles. As the emotional burden of alternatives, since the policy intends to guide political confrontation becomes increasingly future decisions, to be repeated from time to intense, the "available vote" in that society time. The policy establishes a rule to be used tends to move toward them as well, and the for future sets, concerning the same criteria. construction of alternatives different to the Each decision comprises ordering of extremes becomes increasingly difficult. The alternatives (as a gamma problematic). The democratic theory assumes that citizens objective of the study was to provide for one have the ability to be informed, at least decision maker a deeper understanding of partially, and act based on such information, the policy defined by the group and, or at least act as if they are well informed. In furthermore, carry out a comparison different studies it has been observed that between his own beliefs, regarding the political polarization has determinants that criteria and the policy impacts. Those usually respond to historical, socio-cultural, impacts were estimated using realistic economic and geographical phenomena, simulated data, defined by the decision among others. The electoral systems, for the maker. The methodology used is the most part, do not favor the median voter and FITradeoff (Flexible and Interactive trade-off) favor political polarization, since they face because, as an aggregation model, it has the elector with a "large package" of proximity to the defined policy structure. elements that he has to decide, represented Since there is a confidentiality issue, the in a single candidate. The proposal of this focus is on the decision maker/facilitator research is to define an agreement on the interaction. The present work introduces the fundamental problems of a society: violence, use of FITradeoff as a mirror to a decision insecurity, lack of justice, poverty, maker, regarding his own decisions, the employment, health, housing and education, group consensus and its estimated impacts, among many others, and that function as so as to present confirmations of his decision criteria. This agreement is made expectations, but also depicting the lack of through AHP to define the weight of each harmony in some aspects. identified criterion. In addition to this, a

87

3. Evaluation of Energy Efficiency Ana Paula Costa, Universidade Federal Services in Food Industries with de Pernambuco, Brazil, Partial Information based on the [email protected] FITradeoff Method Location problems consist of determining the proper placement of an infrastructure Jorge Augusto Pessatto Mondadori, component in an area of interest while taking SENAI, Brazil, the preferences of decision makers (DMs) [email protected] and existing constraints into account. Mischel Carmen Neyra Belderrain, ITA, Different methodologies have been used to Brazil, [email protected] support location problems, such as the *Rodrigo Ferreira, CDSID/UFPE, Brazil, maximal covering location problem (MCLP), [email protected] Game Theory, Geographic Information Adiel Teixeira de Almeida, Federal Systems (GIS), Multiobjective Optimization University of Pernambuco, Brazil, and Multicriteria Decision-Making Methods [email protected] (MCDM). In public security problems, selecting an appropriate location for a police A decision problem to select the most station considering economic, social, desired energy efficiency service in food criminal and political issues is still an open industries in the south of Brazil was field of research. These issues may be evaluated. The problem was structured represented by different (and somehow through the Value Focused Brainstorming conflicting) criteria, depending on DMs’ methodology and some criteria related to different operational and strategic points of processes, management, knowledge, quality view. This paper proposes the use of the of energy and equipment were defined. The FITradeoff elicitation procedure to support a evaluated alternatives were consulting DM in selecting a specific area/region to services in optimization, diagnostics, locate a new State Military Police Station industrial automation, maintenance, (SMPS) in the State of Pernambuco, Brazil. regulatory consulting and education. For the The State of Pernambuco, located in the preference modeling phase, the use of Northeast region of Brazil, has over 9 million partial information becomes an attractive inhabitants and a territorial area of 98,000 procedure because of the cognitive simplicity square kilometers. The State of when compared to the use of complete Pernambuco’s security system is organized information. In this sense, the FITradeoff into 26 Integrated Security Areas (ISA) that method was analyzed. After structuring the may incorporate one or more districts. This problem together, decision-makers used the arrangement was proposed in 2007 when method individually to identify the priority the then State Government implemented a service to be hired. This method proved security program called “Pact of Life”. This effective in identifying the criteria weights program is associated with a public security each decision maker and indicate the best policy, the main objective of which is to performing alternatives. reduce intentional lethal violent crimes (ILVC), comprising murders and other crimes 4. A Multicriteria Decision Model for that could result in death such as bodily Selecting an Integrated Security Area injury and robberies. The location problem to Allocate a State Military Police behind this paper encompasses two Station based on FITradeoff Method decision-making steps to support the strategic objective (which is, to minimize *Suzana Daher, Universidade Federal crime rates) of the State of Pernambuco’s de Pernambuco, Brazil, Social Defense Department: (1) to define [email protected] which ISA will receive a new SMPS; and (2) to Lucio Silva, Universidade Federal de define a specific geography location (site) Pernambuco, Brazil, [email protected] within this ISA in which the new SMPS will be Katarina Santiago, Universidade installed. This study focused on implementing a model to support the first Federal de Pernambuco, Brazil, decision-making step and on indicating some [email protected] relevant concerns that should be considered for the second step. The contribution of this

88 study to existing literature is two-fold. First, to objective linear programming problems to support DMs in determining which ISA compute the deterioration and improvement should receive a new SMPS using a values of each input/output variable in order multicriteria approach based on an to keep the ISES unchanged. We apply our interactive and flexible tradeoff method, and proposed models for a simple numerical secondly, to provide to public policy example and compute changes for each developers a well-structured model to inputs/outputs variable separately, by only support making decisions on these kinds of solving a mixed-integer linear programming problems. problem. TUE-3-D 2. An Integrated MCDM Approach for Contributed Session: Data Envelopment Evaluating Performance of Airline Analysis Companies Tuesday 13:50-15:30 - Room: Topkapı Melis Almula Karadayı, Istanbul Medipol Palace University, Turkey, [email protected] Chair: Özgür Kabak Umut Aydın, Bandırma Onyedi Eylül 1. Interval Scale Data in DEA: Unıversity, Turkey, Efficiency and Robustness Analysis [email protected] *Füsun Ülengin, Sabancı University, *Akram Dehnokhalaji, Aston University, Turkey, [email protected] United Kingdom, Burç Ülengin, Istanbul Technical [email protected] University, Turkey, [email protected] Nasim Nasrabadi, University of Birjand, Iran, [email protected] Strategic decisions about improving airline Pekka Korhonen, Aalto University, performance have become crucial due to Finland, [email protected] increased aggressive competition in the airline industry. Evaluating airline efficiency Jyrki Wallenius, Aalto University, is an extremely complex, multi-dimensional Finland, [email protected] problem which requires the application of Conventional Data Envelopment Analysis Multiple Criteria Decision Making (MCDM) (DEA) models are developed for the ratio methods. This study evaluates the scale data and they can not provide performance of 45 airline companies via meaningful results for evaluating decision combining the balanced score card (BSC) making units with interval scale data (Halme approach and the network-based super- et al. (2002)). We considered an efficiency efficient data envelopment analysis (DEA). analysis to incorporate interval scale data The proposed methodology incorporates and formulated a mixed-integer linear finance, customers, internal processes, programming model to compute the learning and growth dimensions of BSC into minimum number of units that should be the analysis in order to conduct a eliminated from the Production Possibility comprehensive assessment of airline Set (PPS) in order to make the unit under companies from financial and non-financial assessment efficient. We defined the perspectives of performance. Moreover, the Interval Scale Efficiency Score (ISES) of the eigenvector centrality concept is employed to unit under evaluation as the number of the determine which airlines should act as a role remaining units to the number of all units. model (peer) for efficiency in each dimension We prove that the ISES is a well-defined of BSC. Rankings of airline companies in efficiency measure according to Aparacio each dimension are also presented using the and Pastor’s criteria (Aparacio and Pastor eigenvector centrality values. (2013)). This means that ISES is between 0 and 1, equal to one for only efficient units, strongly monotonic and unit and translation invariant. We also characterize how robust the ISESs are with respect to improvement and deterioration of each inputs/outputs variable. We formulate a general multi- 89

3. Efficiency of Childbirth Services in are reestimated. After this iteration, 75 Turkey: A Jackknifing for Robustness provinces remain for the second model. The same procedure applied for model two and Check of DEA Scores three and finally, recalculated efficiency *Songul Cinaroglu, Hacettepe scores of the final (fourth) model for 72 University, Turkey, provinces show that 13 provinces are [email protected] efficient and 59 are inefficient. The iteration procedure for the final model shows that The motivation of this study is to analyze the reestimated efficiency scores of the final geographic distribution of efficiency of (fourth) model have high correlations with 13 childbirth services in Turkish provinces. Data iterations. In other words, incorporating was collected from official statistical records efficient DMUs into the final (fourth) model of the 2017 Public Hospitals Statistical does not affect the efficiency frontier. The Yearbook. A total of 81 provinces in Turkey, average efficiency score for the final model is representing seven geographic regions, were 0.71. In this study, four different DEA models included in the analysis. Charnes, Cooper, were constructed, and the final model's and Rhodes' input-oriented data efficiency scores were recorded. Finally, a envelopment analysis (DEA) was applied to decision-tree procedure was integrated into determine provinces' efficiency scores, using the DEA results, and predictors of efficient childbirth-specific input and output and inefficient provinces were examined. The indicators. Input variables of this study are as study's results showed that 18% of provinces follows: total number of physicians', total were efficient in terms of childbirth services. number of nurses and midwives' and number Average efficiency scores were high (0.71) of beds in neonatal intensive care units. for provinces located in the Southeast Output variables are: number of normal Anatolia region of Turkey. Decision tree delivery, C-section and operative delivery. A shows that Number of beds in neonatal jackknifing was applied for a robustness intensive care units (Neo_int_n_b) is the check of DEA scores. This procedure is most important predictor of efficiency for performed to test the robustness of DEA childbirth services. Geographic distribution efficiency scores, obtained from input and of childbirth services' efficiency scores in output variables in 81 Turkish provinces. The provinces shows that Eastern Turkey has the iteration procedure generates four models. highest score. Neo_int_n_b, one of the input The first, called the “prior model,” has 81 indicators under the control of health DMUs. According to this model, 22% of professionals to better manage childbirth provinces are efficient. (In this model, 18 services, is the most important determinant DMUS are efficient, and 63 are inefficient). of efficiency scores. Ensuring public health The average efficiency score for the first managers' awareness and continuous model is 0.69.18 iterations were applied for monitoring of childbirth services, while also prior model for every efficient DMU. During focusing more on regional differences, are the prior model's jackknifing iteration essential to improve the status of children’s procedure, efficiency scores from the first six health in Turkey. iterations presented low correlations with the prior model. In other words, reestimated 4. DEA with Window Analysis and efficiency scores of provinces, after dropping Cluster Analysis Approach for the six most-efficient provinces one at a time Analyzing the Interaction of Human from the analysis, indicate a low level of Development and Competitiveness correlation with the prior model. Clearly, efficiency scores obtained without including *Hakan Kılıç, Koç University, Turkey, the six most-efficient DMUs one at a time [email protected] does not give results similar to the prior Özgür Kabak, Istanbul Technical model’s efficiency scores. Considering the University, Turkey, [email protected] six most-efficient provinces in the prior model affects the efficiency frontier. In this Purpose – Countries utilize their regard, to avoid extreme outliers that could competitiveness to achieve human affect the robustness of efficiency scores, development. On the other hand, when a the six most-efficient DMUs in the prior country advances in their human model are removed, and efficiency scores development, it is likely their

90 competitiveness level also increases. This 1. Bi-Objective Optimization in study aims to analyze this bilateral relation Complex GAS and Liquefied Natural among human development and competitiveness at country level. Gas Transport Network Design/methodology/approach – United Mokrani Amira Nagham, USTHB, Nations Development Programme’s (UNDP) Algeria, [email protected] Human Development Index (HDI) and World Chaabane Djamal, USTHB, Algeria, Economic Forum’s (WEF) Global [email protected] Competitiveness Index (GCI) is used to represent countries’ human development Naturel gas is expected to occupy a and competitiveness respectively. Data preponderate place in the global energy Envelopment Analysis (DEA) with window balance. As regards Algeria, a country that is analysis is employed to compute the a producer, an exporter and a consumer, this countries’ efficiencies on converting their natural resource plays a key role in the human development to competitiveness and country. The discovery and exploitation of the inversely, competitiveness to human deposits, added to that already in operation, development with 3-year time lag. Then, a will increase the production of hydrocarbons. stepwise cluster analysis for time periods are This growth of production will necessarily used to analyze the results of the DEA have direct consequences on the transport, models. 56 countries that comprise the 90% storage and export capacity of liquid of the globe’s population or GDP are hydrocarbons, which will follow a growing evaluated for the years 2010-2017 using the trend in response to market demand to both proposed methods. Findings – No drastic national and international levels. The change is observed as according to the development of a global model integrating results 77% of countries remain in the same the gas pipeline from the liquefaction plants cluster in the investigated period. Input-GCI to the marine terminal does not exist. Hence, Output-HDI model is more reliable than we will develop a multicriteria decision- Input-HDI Output-GCI model, since its results making tool that will act as a standard of are consistent with the cluster analysis. experimentation where we can see the Therefore, the direction of the relation impact of any decision that can be taken on between competitiveness and human the whole chain. In order to solve this development is from competitiveness to problem , first, some hydraulic calculations human development. The competitiveness are performed to transform this problem by level of a country in a certain year affects its writing it as a mathematical model ( linear human development with a time lag. The programming problem). In the second step, basic implication of this result is that the we passe through two stages : in the first countries needing a development in human- stage, we prepare the parameters of the well-being can focus on the improvements on model using the excel solver then, in the their competitiveness. Originality/value – To second stage, the corresponding linear the best of our knowledge, this is the first programming problem is solved using the study to investigate the bilateral relationship IBM ILOG CPLEX 12.6 solver to obtain the between human development and optimal solution. competitiveness. To that end, we use DEA and cluster analysis. Additionally, 2. Decision Making in Pump longitudinal data is incorporated to observe Hydraulic Design the trends. *Tomáš Krátký, Centrum Hydraulického TUE-3-E Výzkumu spol. s r.o., Czech Republic, [email protected] Invited Session: Continuous Tinkle Chugh, University of Exeter, Multiobjective Optimization with United Kingdom, [email protected] Engineering Applications Kaisa Miettinen, University of Jyvaskyla, Tuesday 13:50-15:30 - Room: Hagia Sophia Finland, [email protected] Chair: Michael Stiglmayr Lukáš Zavadil, Centrum Hydraulického Výzkumu spol. s r.o., Czech Republic, [email protected]

91

Vít Doubrava, SIGMA Výzkumný a 3. Biobjective Shape Optimization Vývojový Ustav, s.r.o., Czech Republic, Using Weighted Sums: Probability of [email protected] Failure versus Costs In this study, we want to show the practical *Johanna Schultes, University of role of the decision making in the pump Wuppertal, Germany, hydraulic design, and the challenges connected to it This involves an iterative [email protected] process combining computationally very Kathrin Klamroth, University of expensive numerical simulations and Wuppertal, Germany, analytical formulas with expertise and [email protected] intuition. To obtain an optimal performance, Michael Stiglmayr, University of a careful balancing of the trade-offs between Wuppertal, Germany, various, conflicting objectives, is crucial. The [email protected] real design problem considered is related to Hanno Gottschalk, University of the development a diagonal pump with an Wuppertal, Germany, axial diffuser. In our case, the pump was [email protected] designed to replace an already existing but outdated pump. This setting imposed many We analyze the trade-off between the limitations, and forced compromises in the Probability of Failure of a ceramic component design and its performance. To meet the under tensile load and its material requirements demanded by the customer, it consumption using the weighted sum took several rounds of discussions with the scalarization of a biobjective shape decision maker and corrections based on optimization problem. Numerical numerical simulations. Finally, to further experiments indicate that even though the improve the design, we created a parametric problem is non convex in general, weighted model of the diffuser and performed an sum scalarization is a useful tool in this optimization. In total, there were 22 context due to its numerical robustness. With geometric parameters and three objectives this approach we are able to overcome local (efficiency at three different flow rates). The minima of the biobjective shape optimization numerical simulations were performed in problem and control the trade-off between ANSYS CFX software, and one computation the two objectives while varying the weights. took ca. 20 hours - which limited the number For solving the weighted sum scalarization of objective evaluations. Two different we outline a steepest descent algorithm and surrogate-assisted methods were used. First, apply it to two test cases based on a finite a single-objective method called “Stochastic element discretization using B-splines as RBF”. Scalarization of the objectives was control variables. In the end we also address necessary in this case, the weights were the drawbacks of the weighted sum based on the discussion with the decision scalarization and compare our results to an maker. Next, we decided to employ K-RVEA, alternative approach, a gradient-based an evolutionary surrogate-assisted, biobjective optimization algorithm. multiobjective optimization method which utilizes Kriging to shorten computing times. 4. Gradient-Based Biobjective Shape The K-RVEA displayed a superior Optimization: Probability of Failure performance to Stochastic RBF. Finally, one versus Cost. design was selected, and a model pump was produced and experimentally measured in a *Onur Tanil Doganay, University of hydraulic laboratory. The outcome was Wuppertal, Germany, satisfying, as all required performance [email protected] indicators were met or exceeded. Michael Stiglmayr, University of Wuppertal, Germany, [email protected] Kathrin Klamroth, University of Wuppertal, Germany, [email protected]

92

Hanno Gottschalk, University of TUE-3-F Wuppertal, Germany, [email protected] Contributed Session: Data, Applications and MCDM Shape optimization problems usually involve several conflicting optimization criteria like, Tuesday 13:50-15:30 - Room: Basilica for example, cost, reliability and efficiency of Cistern the shape. In most engineering applications, Chair: Seda Yanık these optimization criteria are incorporated in one single-objective optimization problem 1. Implementation of an Enterprise- by either using weighted sums scalarizations Level Open Source APM System and or by introducing constraints on, for example, Comparisons with Alternatives Using the maximum allowable cost or weight of a the MCDM Methods component. In this study, we suggest a truly multiobjective optimization approach to *Ali Özmez, Doğuş Teknoloji, Turkey, shape optimization using the two [email protected] optimization criteria reliability and cost, and Can Alptekin, Doğuş Teknoloji, Turkey, present a numerical method to approximate [email protected] the Pareto critical set for realistic test cases. Since the necessary discretization induces a Application Performance Monitoring (APM) high computational cost, we suggest to apply systems are used for monitoring and derivative-based methods and adapt a management of performance and availability multiobjective descent algorithm introduced of software applications. APM systems strive by Fliege and Svaiter in 2000. While this to detect and diagnose complex application method was originally developed to compute performance problems to maintain an one Pareto critical point, we suggest a expected level of service. The objective of combined scaling and reference point this research is to find the best possible open approach to obtain several meaningful source application or tool with various R&D solution alternatives that approximate a techniques that can help achieve the goal of Pareto critical front. Besides the creating the most efficient and complete identification of an efficient multiobjective APM system by making comparisons with optimization tool, this involves the derivation existing alternatives using the MCDM of differentiable objective functions Methods. By putting together various APM representing the reliability and the cost of a tools, it is aimed to achieve a complete APM component. While sensitivities for cost and solution to address the needs of an fluid dynamic objective functions have been organization for monitoring its applications used for a long time, this is not the case for on user-friendly dashboards. We show how objectives related to reliability since peak we can efficiently combine the several open stress is not a differentiable function of the source tools such as Elastic Stack, shape. This changes, if deterministic Prometheus, Libbeat, Zabbix, and Grafana to reliability criteria are replaced by a implement an APM system. Elastic Stack, probabilistic criterion, namely the probability Prometheus, Libbeat, Zabbix, and Grafana of failure (PoF), which has been introduced are important open source tools that can be recently to the field of ceramic design and used to implement an APM (Application low cycle fatigue. Here we present a finite Performance Monitoring) system when put element based first discretize, then adjoin together. Built on an open source foundation, approach for the calculation of shape the Elastic Stack (formerly ELK Stack) lets gradients (sensitivities) with regard to the you reliably and securely take data from any PoF of ceramic designs. This is applied to the source, in any format, and search, analyze, simultaneous minimization of the PoF and and visualize it in real time. Prometheus the volume of a 2D ceramic rod in a collects metrics from monitored targets by biobjective shape optimization problem. We scraping metrics HTTP endpoints on these show numerical results for two test cases targets. Grafana allows you to query, and discuss the validity of the resulting visualize, alert on and understand your shapes. metrics no matter where they are stored. Using Grafana or Kibana, you can create, explore, and share dashboards with your

93 team and foster a data driven culture. The implementation of any analysis or Data Libbeat library written entirely in Go, offers mining of spatial data requires the building an API that Beats can use to send data to and structuring of a Spatial Data Warehouse Elasticsearch, while enabling gathering IIS (SDW). This SDW must be readily usable by logs, audit logs and many other properties. GIS and by tools proposed by an OLAP Finally, Zabbix provides a solution to gather system. This work aims to develop MCDM various types of real-time monitoring data and SDW methods which will be integrated from tens of thousands of servers, virtual into a GIS according to a "GIS-dominating" machines, and network devices for the approach. The functioning tools of GIS will be purpose of alerting. We designed an APM operational to exploit the SDW. system based on the above given tools using the publicly available plugins to integrate 3. A Data-Driven Multi-Criteria them. For this we used Prometheus to collect Optimization Framework for Bus metrics, Elastic Stack to collect logs, Grafana Scheduling in Public Transportation and Kibana to create dashboards, Libbeat to send monitoring data to Elasticsearch, and Seda Yanık, Istanbul Technical Zabbix to gather real-time monitoring data University, Turkey, [email protected] from all the system devices. Finally, utilizing *Salim Yılmaz, Istanbul Technical the MCDM Methods, we compared the University, Turkey, designed APM system with other solutions in [email protected] the open source world having the same purposes (e.g. OpenAPM). As the vast majority of the world’s population has started to live in cities, issues such as 2. Implementation of Spatial Data increased mobility needs, traffic congestion Warehouse and Multi-Criteria and carbon emissions are becoming Decision Making in GIS important. In addition to all of these, the changing dynamics in the markets push the *Abdellah Mebrek, Centre des citizens into the expectation of techniques Spatiales, Algeria, transportation service in a comfortable and [email protected] high quality manner. Thus, the companies Ahmed Saidi, Centre des Techniques providing mobility and the researchers are Spatiales, Algeria, [email protected] required to produce sustainable transportation solutions considering these Zohra Makranfar, Centre des issues. To provide efficient planning Techniques Spatiales, Algeria, solutions for the transportation in cities, we [email protected] deal with the scheduling problem of the GIS offers advanced functions for buses in the public transportation. In large acquisition, storage, analysis and display of metropolitan cities, the bus network is very geographic information. However, their large. Thus, planning of services is difficult effectiveness for complex spatial analysis is due to changing demand throughout the day, questionable because of their determinism traffic congestions and the need to use the and decision-making rigor. The MCDM scarce resources optimally. In this study, we methods can make some solutions for a set formulate an integer-programming for the of problems with various and multiple optimization of a bus line. The problem seeks criteria. When the problem is so complex it is for the solution of optimal frequencies of a judicious to combine MCDM process with specific bus line’s various short turn and other approaches like Data mining. The OLAP complete trip alternatives that are used to technology which combines both bases balance the demand differences between multidimensional analysis and the concepts the links of the line. We consider two of the Data mining provides powerful tools objectives in our model formulation: (i) allowing the highlighting inductions and minimization of the difference between the information not obvious by the traditional provided service capacity and the demand; tools. However, these OLAP tools become and (ii) passenger travel time. In order to more complex in the presence of the spatial reduce these two objectives into a single dimension. The integration of OLAP with a objective, we obtain the meaningful weights GIS is the future solution for geographic and for each objective using the Analytical spatial information. A precondition for the Hierarchy Process (AHP). The first objective,

94 namely difference between the provided areas in the traditional grocery distribution service capacity and the demand, is such as reducing the carbon emissions by evaluated in terms of the criteria of operating better routing, changing the transportation cost minimization from the operator modes or eliminating the food spoil during perspective and comfort from the passenger the distribution. In this study, we propose a perspective. The second objective, new urban distribution model for grocery passenger travel time is evaluated in terms goods particularly by integrating the railway of the criteria of customer satisfaction from mode in the distribution network and the operator perspective and utility from the enhancing the last mile delivery via passenger perspective. Moreover, we use a collaboration with the small independent clustering approach to cluster the time of day grocery shops dispersed densely around the based on the demand. Thus, we obtain the urban area. We assume that there are cutting hours of the morning peak, off-peak candidate grocery points within a given and evening-peak hours and find the optimal residential area and that the demand of the frequencies for these different clusters of the customers can be divided. The problem is time-of-day separately. We present a case defined as a multi-objective maximal study in Istanbul which is one of the largest covering hierarchical facility location cities of the world with a population of 15 problem. The hierarchical facility network million. An extensive dataset is obtained considered in this problem consists of three from IETT that provides the bus public levels, that are (i) railway stations, (ii) small transportation service in Istanbul. The grocery stores and (iii) end-customers. It is dataset that we use is related to one of the assumed that the railway is used to replenish longest lines of Istanbul with the line the small grocery stores from a grocery number, 500T. The bus line, 500T has warehouse which is close to the starting almost 70 stops and travels through highly point of the railway line. The grocery goods congested links throughout the day. are sent to from this warehouse to the Moreover, the buses are commonly over- specific railway stations to which grocery crowded. To overcome these problems, we stores are assigned to and the grocery goods use mathematical programming approach to are picked up by the grocery stores. Contrary specify optimal short-turn and complete trips to the common practice of using trucks for and their frequencies. Using the real-life the long-distance distribution process, using data, we implement our model and solve it rail line as proposed in this study would using CPLEX. We specify the weights of the contribute to cost-effectiveness and the two objectives by AHP using expert opinion. carbon emission reduction even though, the We also present the pareto-optimal latter is not explicitly added in the solutions. mathematical model. The network is also designed by incorporating the small grocery 4. A Data-Driven Optimization Model stores in order to maximize the coverage of for the Urban Grocery Distribution the end-customers and to minimize the Network collecting cost of goods from the railway stations. The fulfilment of the end-customer *Seda Yanık, Istanbul Technical orders is assumed to be either done by the University, Turkey, [email protected] grocery store on foot or by the end-customer Sibel Özkardeş, Kocaeli University, and is not considered in the mathematical Turkey, [email protected] model. We formulize the problem as a three objective mixed integer programming model. Grocery distribution is commonly multi- The first objective is to maximize the layered starting from the farmer to the fulfilment of the demand by ensuring that wholesaler, then to logistics service provider, customers receive the service from the to the retailer and finally to the end- nearest grocery stores with the specified customer. Thus, an effective and sustainable coverage parameter. The second is to distribution network design considering the minimize the distance between railway whole chain is quite complex and stations and grocery stores. The third is to challenging. With the increasing use of equalize the income of all the grocery stores internet and e-tailing, last-mile delivery has by determining which grocery store meets also become one of the challenges of grocery the demand of the customers. We propose a distribution. There exist many improvement data-driven optimization approach for a case

95 study problem in Istanbul, Turkey. The urban effect of the method of normalization on the railway to be used is a metro line named as ranking of alternatives for MCDM rank “Marmaray”. Marmaray line goes through the methods. The rank model MCDM has the urban area from one end to the other and is following form: Q=f(A, C, DM, ’w’, ‘nm’, ’dm’), assumed to be utilized for goods and includes the choice of the set of transportation before starting the daily alternatives (A) and the set of criteria (C), the scheduled earliest passenger transportation estimation of the values of the attributes of trip. To analyse the grocery demand in the alternatives (DM), the evaluation of the Istanbul with respect to the demographical weight of the criteria ('w'), the choice of the and distance based variables, we use a normalization method (΄nm΄) of the decision credit card transaction data in the category matrix, the choice of the metric to calculate of grocery from a large Turkish bank. Dataset the distances in n-dimensional space of includes the distance between the customer criteria (΄dm΄), the choice of aggregation shopping location and the location of the method (f) of attributes of alternatives. In customer base (i.e. either the home and work accordance with the method of aggregating location) as well as the demographic the attributes of alternatives, the variables together with the three-month assessment score Q of each alternative is customer grocery spending data. After the formed and ranking is made. For different data analysis, we incorporate the obtained normalization methods (΄nm΄), the information from the dataset to the assessment score of alternatives Q can vary mathematical programming model and solve significantly. The difference consists not only it using Cplex. We further analyse the effect in absolute values, but also in relative values, of the coverage parameters and the number which changes the ranks of alternatives of grocery shops in the network by comparing within the framework of the applied the results of the numerical studies. aggregation method. The author presents a new method for normalizing the decision TUE-4-A matrix (IZ-method) for solving multi-criteria Contributed Session: Multiple Criteria decision-making problems. Normalization of Ranking and Sorting Methodology the elements of the decision matrix is performed in such a way that: I. To preserve Tuesday 16:00-17:40 - Room: Galata Tower the proportions of natural and normalized attribute values of alternatives for each Chair: Marko Bohanec criterion – the principle of "vertical" 1. A New Method for the normalization. II. To equalized the largest Normalization of Multidimensional and smallest values of the attributes of alternatives for all criteria – the principle of Data (IZ-Method) for MCDM "horizontal" normalization. This ensures the Problems equality of the contribution (on average) of *Irik Mukhametzyanov, Ufa State each criterion to the assessment score and the absence of priority of individual criteria Petroleum Technological University, before assigning weights. IZ-method is a Russia, [email protected] multi-step method. The first step is the A review and critical analysis of currently normalization of the attributes of alternatives methods for the normalization of using the Max method. Next, the largest and multidimensional data in problems of multi- smallest attribute values of the alternatives criteria decision making is presented. It is are aligned, the sorting of values of shown that within the framework of the alternatives for cost criteria is performed, existing methods of normalization the and then the normalized values are problems of scaling and converting scales, “stretched and compressed” in proportion to the asymmetry of normalized values, the the natural values. As a result of such conversion of cost criteria into the benefit transformations, the normalized values criteria are not solved. There are no criteria retain the most complete information of the for the effectiveness of linear and nonlinear natural values of the attributes of the normalization methods. The main part of the alternatives. The advantage of the IZ- statements is given using formal method: – IZ-method allows scaling mathematical proofs. The focus of the study operation of natural values of attributes of is on the problems of normalization and the alternatives based on linear transformations;

96

– IZ-method excludes the priority of the which takes into account all the DM’s contributions of individual criteria to the preferences given in the form of pair-wise assessment score of the alternative; – IZ- comparisons. The underlying theory of the method solves the problem of the asymmetry convex cone based models is as follows. of normalized values; – IZ-method solves the Assuming that the DM has presented his/her problem of conversion the cost criteria in the preferences in the form of pair-wise benefit criteria; – using the nonlinear method comparisons, some preference subsets are at the first step of the IZ-algorithm gives formed, consisting a subset of alternatives better results of proportionality of the along with a single alternative which is attributes of alternatives for each criterion, indeed the worst one among those. Then compared to the procedure of ordinary associated with each preference subset, a nonlinear normalization. The report presents convex preference cone and a preference the full algorithm of the IZ-method and a polyhedron are defined. Finally, considering computer program in the MatLab System. each preference subset and for each Numerous comparative numerical examples alternative does not belong to it a simple are performed for 8 basic normalization linear programming problem checks whether methods and 15 aggregating methods for this alternative is dominated by the cone or the attributes of alternatives when not. In case it is cone-dominated, it is calculating the assessment score of each eliminated from further consideration. This alternative. simple and straightforward technique has led to several MCDM approaches for 2. A Review of Using Convex different goals, such as finding the most Preference Cones in Multiple Criteria preferred solution, partial and total ranking Decision Making of alternatives, evolutionary algorithms to find a part of Pareto optimal frontier, etc. In *Nasim Nasrabadi, University of this paper, we provide a review on using Birjand, Iran, convex preference cones in MCDM. [email protected] References: Korhonen P, Wallenius J, Zionts Akram Dehnokhalaji, Aston University, S (1984) Solving the discrete multiple United Kingdom, criteria problem using convex cones, [email protected] Management Science 30 (11): 1336–1345. Pekka Korhonen, Aalto University, Korhonen P, Moskowitz H, Wallenius J Finland, [email protected] (1986) A progressive algorithm for modeling Jyrki Wallenius, Aalto University, and solving multiple-criteria decision Finland, [email protected] problems, Operations Research 34 (5): 726- 731. Multiple Criteria Decision Making (MCDM) is a branch of operations research dealing with 3. Activity Based Project Risk optimization problems concerning a number Assessment Method Development of conflicting objectives and criteria. Based via AHP-Stochastic TOPSISs Hybrid on the orientation of different MCDM problems, several methods have been Algorithm proposed to fulfill the underlying goals. One *Emin Başar Baylan, Istanbul Ticaret of the most well-known methods in the field University, Turkey, of MCDM is based on using convex [email protected] preference cones, originally developed by Korhonen et al (1984). The main assumption In project planning, risk assessment method in implementing convex cones is that the plays vital role. Poorly assessed project risks DM's value function is increasing and quasi- cause degeneration at project cost, project concave, provided that all objectives are in completion time, and project output quality the form of maximization. A simple approach and project scope. Each project activity risk developed by Korhonen et al. (1986) can be influence these project success factors. implemented for checking that whether there Implementation performance of a project exist a quasi-concave value function activity triggers or smooths of its successor’s consistent with the Decision Maker (DM)’s activity risks. Because of this; employing preferences. Their method is based on a robust and detailed risk assessment simple a simple linear programming problem methods is important to reach those project

97 goals. In project risk assessment literature, possible by employing the principle of when it is investigated, it is noticed that risk dominance, which generally yields a partial assessment and evaluation methods are ranking. This can be further extended by only developed at whole project level. constructing an additional quantitative Actually, they are not comprehensive enough (numerical) evaluation model aimed at total to evaluate the project risks at activity level. ranking. Such model should be constructed Besides that traditional risk assessment automatically using only information that is methods such as risk matrix does not able already contained in the qualitative DEX analyse project risk quantitatively. With this model, without requesting any further motivation, main aim of this study is information from the decision maker. The developing a multi-criteria based decision quantitative model has to fulfill several method which prioritizing project risks at requirements: 1. Consistency of qualitative activity level. AHP and TOPSIS method are and quantitative evaluations: For each combined to developed novel method. In this alternative assigned to class ‘C’, the hybrid method, Constructing AHP model is to corresponding numerical evaluation must lie prioritize work packages with respect to in the interval [c – 0.5, c + 0.5], where ‘c’ is relative importance of project time, project the ordinal number of ‘C’. The values c + 0.5 output quality and project cost. Broken down and c – 0.5 are interpreted as “ideal” and structure of these work packages are used “anti-ideal” evaluations within ‘C’, as input for weighted criteria for TOPSIS respectively. 2. Compatibility of inputs and method. In second layer of this decision outputs: DEX models are hierarchical, method, TOPSIS model is used for prioritizing therefore evaluations obtained as outputs at predetermined activity risks according model subtrees enter as inputs to the higher weighted project work packages success levels of the hierarchy. Thus, the quantitative criteria. In the application of this method, a scales of input and output attributes must be case study approach is followed. In this compatible and should obey the “c +– 0.5” sense, “Global Furniture Ltd.” which is principle. 3. Quantitative evaluations must established in Istanbul, Turkey is chosen as preserve the dominance between a case to apply newly developed model. alternatives. So far, two such methods have Results showed that application of AHP- been used in connection with DEX. They Stochastic TOPSIS Hybrid Algorithm provides interpret decision rules as points in a multi- a platform that project risks could be dimensional space and try to approximate analysed as quantitative and also at project them with some suitable numerical function. activity level. An older method, called QQ (Qualitative- Quantitative), constructs per-class 4. Improving Within-Class Ranking of hyperplanes, and a more recent method Decision Rules in Qualitative Multi- constructs per-class copula functions. Both Criteria Method DEX methods have been successfully used in practice for ranking research projects, *Marko Bohanec, Jožef Stefan Institute, commutator motors and subject-predicate- Slovenia, [email protected] object triplets in exploring literature DEX is a qualitative multi-criteria method, in domains. This work explores the idea to which all criteria are represented by improve within-class discrimination of qualitative (symbolic, verbal) attributes. The alternatives by formulating the construction attributes are structured into a hierarchy, of quantitative evaluation models as an and the evaluation of alternatives is optimization problem. In a nutshell, all governed by decision rules, defined by the decision rules that map an alternative to decision maker. DEX is particularly suitable class ‘C’, should be distributed within the [c for “sorting” decision alternatives into – 0.5, c + 0.5] interval so as to maximize the distinct preferentially ordered classes. The distances between the dominating- method DEX is implemented in the software dominated pairs of decision rules while trying DEXi to satisfy the above requirements. Following (http://kt.ijs.si/MarkoBohanec/dexi.html). this idea, four new methods have been In DEX, all alternatives assigned to some designed: two based on linear optimization, class ‘C’ are mutually indifferent from each one on quadratic optimization and one that other. However, it is often desired to further algorithmically explores the within-class rank alternatives within classes. This is dominance graphs. These methods were 98 experimentally assessed and compared with 2. An ANP-based Methodology to QQ on 3322 DEX real-life decision tables. All Evaluate Energy Strategies in Turkey the new methods significantly outperform QQ in terms of within-class discrimination of Dursun Delen, Oklahoma State decision rules. Among themselves, the new University, United States, methods give very similar results, with no [email protected] clear winner. Any of those methods will *Beyzanur Cayir Ervural, Konya Food improve the within-class evaluation of and Agriculture University, Turkey, alternatives in future versions of DEX. [email protected] TUE-4-B Selim Zaim, Istanbul Sehir University, Turkey, [email protected] Contributed Session: Practical Applications of AHP/ANP Planning of energy strategies is a complex decision-making process that encompasses Tuesday 16:00-17:40 - Room: Maiden's a wide variety of interrelated sub-processes Tower and related criteria where the ultimate goal Chair: Sarbast Moslem is to meet the variant needs and wants of all stakeholders. Due to energy vulnerability at 1. Performance Assessment of the global scale, its critical and sensitive Transportation Options nature, countries need to evaluate their energy policies comprehensively, rationally, Mine Isik, Bogazici University, Turkey, and accurately. In order to adapt to the ever- [email protected] changing energy market conditions, *Özay Özaydın, Doğuş University, governments are in need of constantly re- Turkey, [email protected] planning their energy policies, taking into account their short-term and long-term One of the crucial elements of human needs strategic policies. Because of its multi- is energy. Our energy choices and decisions faceted nature, the determination of the right have a substantial impact on natural energy policy is deemed to be a complex yet systems. As population, economic growth, crucial managerial problem. In this study, we and urbanization increases globally, energy employed an Analytic Network Process demand increases even at a higher pace. By (ANP)-based methodology, which is one of definition transportation is the movement of the most promising and, as per the recent humans, animals and goods from one literature in this and similar application location to another thus intertwined with domains, most effective multi-criteria energy. Within the industries’ demanding decision-making (MCDM) tools. ANP is energy, transport plays a crucial role in capable of addressing the complexities exacerbated mobility. Transportation sector imposed by this decision-making process that induced by high population and because of its comprehensive working industrialization emits a significant portion of principle coupled with its network structure total emissions. By all available modes for that can take into account the interaction not only individual but also collective from various elements at different levels. passenger transportation (subway, suburban Since ANP can capture multi-dimensional train, bus, passenger cars, bicycles, etc.), relationships between alternatives and meeting this steadily increasing travel criteria, it helps in producing objective and demand and lowering its impact on the comprehensive results. In this study, all environment has been a hot research topic, factors and subfactors are determined as no standard solution has been found. As according to the Strategic Plan of the sustainable mobility planning consists of Ministry of Energy and Natural Resources many dependencies and links among factors and evaluated using the Strengths, influencing the decision-making process, it Weaknesses, Opportunities, and Threats has a perfect fit to use Multiple Criteria (SWOT) analysis under the general theme of Decision Making (MCDM) methodologies, strengths/weakness and and this paper presents an assessment of opportunities/threats of the energy market. environmental impact preferences of In this context, four main criteria (i.e., transport. themes), 24 sub-criteria, and nine alternatives are defined. The weights are 99 calculated and energy strategy priorities—to occupancy) and social (e.g., social reveal the importance levels from high to reputation, workers’ wellbeing) criteria. low—are determined using the ANP-based MCDM methodology. 4. Enumerating the Difference of Preferences Between an Emerging 3. Ranking of Seismic Retrofit and Developed City Related to Public Strategies for Industrial Buildings: A Transport by Using the Analytic Hierarchical Approach Hierarchal Process Chiara D'Alpaos, DICEA - University of *Sarbast Moslem, Budapest University Padova, Italy, [email protected] of Technology and Economics, Hungary, *Francesca Andreolli, DICEA - University [email protected] of Padova, Italy, Szabolcs Duleba, Budapest University of [email protected] Technology and Economics (BME), Flora Faleschini, DICEA - University of Hungary, [email protected] Padova, Italy, [email protected] Public transport improvement has always Recent destructive earthquakes in Northern been a crucial issue to decision makers as and Central Italy resulted in considerable well as to transport experts, especially in damages to real estate with respect to emerging and developed cities. The aim of residential and industrial buildings and this study is to evaluate public transport historic centers. Earthquakes occurred in demand between an emerging city “Mersin” May 2012 in Emilia-Romagna caused extent and a developed city “Budapest”. As a and serious damages to the economy of this methodology, the analytic hierarchy process area, due to the collapse of a large number approach has been proposed to rate the of reinforced concrete (RC) precast industrial most critical public transportation supply buildings, characterized by high seismic quality criteria from transportation experts vulnerability. The estimated direct and point of view, the data has been collected by indirect economic costs were about 1 and 5 applying pairwise comparison questionnaire billion Euros, respectively. In this context, the survey. AHP based on the dynamic analysis management of post-earthquake recovery and sensitivity analysis gives the decision operations and the adoption of preventive maker the confidence of the consistency and seismic risk mitigation strategies is of the robustness. Rating the criteria in the paramount importance. Effective strategies hierarchical structure is comprehensive, require the stabilization of severely damaged flexible and shows a great potential for buildings and the preventive improvement of supporting decision-makers with the constructions structural response to seismic transportation decision-making process. The actions. Although the operation of two different groups showed the different emergency inspections are meant to classify views of development between decision buildings on the basis of building residual makers and the public; this might be due to seismic capacity, decisions on prioritization many factors like costs and political situation of interventions depend on different criteria in the area that make public bus related to both buildings structural transportation development plans logical. characteristics, materials and geometrical The paper provides recommendations to properties, and to peculiarities of enhance the operation for the general public manufacturing and production of goods and transport system. services made. In this paper, we investigate different seismic retrofit strategies and we TUE-4-C propose an AHP model for multi-criteria prioritization of post-earthquake intervention Contributed Session: Interactive on industrial buildings, damaged by Multiobjective Optimization earthquakes occurred in North-Central Italy Tuesday 16:00-17:40 - Room: Dolmabahçe in the last decade. We take into account Palace multiple criteria grouped into technical (e.g., vulnerability, compatibility, reversibility), Chair: Kaisa Miettinen economic (e.g., construction costs, costs due to production downtime and disruption of

100

1. Interactive Procedure for 2. Focused NAUTILUS - An Multiobjective Dynamic Interactive Method Parallelizing the Programming with the Mixed Human and the Machine Ordered Structure *Seyed Mohsen Mousavi, University of Maciej Nowak, University of Economics Jyväskyla, Finland, [email protected] in Katowice, Poland, Dmitry Podkopaev, Polish Academy of [email protected] Sciences, Poland, Sebastian Sitarz, University of Silesia in [email protected] Katowice, Poland, Kaisa Miettinen, University of Jyvaskyla, [email protected] Finland, [email protected] *Tadeusz Trzaskalik, University of Many multiobjective optimization methods Economics in Katowice, Poland, are interactive by nature. The solution [email protected] process is divided into iterations, where in Multiobjective, multistage decision problems each iteration the DM provides some are usually investigated as models of preference information to the method, then multiobjective dynamic programming, using the method derives Pareto optimal solutions the vector version of Bellman’s principle of corresponding to these preferences and optimality non-dominated evaluations (in the provides to the DM new information about criteria space) and efficient solutions (in the the Pareto optimal set. Interactive methods decision space). Deterministic, stochastic have many applications in business and and fuzzy models are examples or this industry. One of their benefits is the efficient approach. Another method of generalization utilization of computational resources, since of single-criterion dynamic programming the method derives in each iteration only consists in regarding the evaluations as those Pareto optimal solutions, which elements of a partially ordered space. A correspond to the preference information. problem that appears in many decision Solving real-life problems can involve heavy models is that of the simultaneous computations or simulations. Even deriving occurrence of deterministic, stochastic, and one Pareto optimal solution may result in an fuzzy values in the set of multidimensional unacceptably long waiting time for the DM. evaluations. A question arises: Can such One approach to overcoming this issue is mixed evaluations be used in optimal control using an approximate, finite representation of a multiobjective decision process of the Pareto optimal set constructed by according to a homogeneous scheme in means of a population-based heuristic ordered structures? The paper presents a algorithm prior to the interactive solution multiobjective dynamic programming process. Then, in each iteration, instead of problem with the values of the criteria solving the original problem, the method function in ordered structures. The first selects an element from the representation problem is a model with deterministic values; best corresponding to DM’s preferences. the second, one with triangular fuzzy This approach allows eliminating waiting numbers; and the third, one with discrete times at the expense of reducing the quality random variables with the k-th absolute of information presented to the DM in each moment finite. The fourth model is a product iteration. In the case of computationally of the three models listed above. The aim of complex problems, the quality of the Pareto the paper is to present an interactive optimal set representation may be procedure which uses trade-offs and which unacceptable due to limited preparation time allows to determine the final solution in the and available resources. In order to mixed ordered structure. The ordered overcome this issue, we propose to hybridize structures and the proposed procedure are an interactive method with a population- illustrated by numerical examples. based heuristic in an intelligent way. Namely, after constructing the initial representation of the Pareto optimal set, the heuristic algorithm continues running in the background in parallel with the interactive solution process. The algorithm constantly improves the quality of representation 101 utilizing the additional running time. 3. Surrogate Assisted Interactive Moreover, the algorithm focuses on those Multiobjective Optimization in areas of the objective space, which are currently interesting for the DM. The smaller Building Energy System Design are the areas, the higher is the quality of *Pouya Aghaei Pour, University of representation which can be achieved in Jyvaskyla, Finland, pouya.p.aghaei- them given the same resources. Our new [email protected] hybrid method is called focused NAUTILUS Tobias Rodemann, Honda Research (F-NAUTILUS). It is based on E-NAUTILUS, the method is designed to work with finite Institute Europe, Germany, representations of the Pareto optimal set [email protected] using the generic NAUTILUS concept. It starts Markus Olhofer, Honda Research from the worst possible objective vector and Institute Europe, Germany, in each iteration, improves all the objective [email protected] values, thus moving towards the Pareto Jussi Hakanen, University of Jyvaskyla, optimal objective vectors. The direction of Finland, [email protected] improvement is set by the DM via selecting Kaisa Miettinen, University of Jyvaskyla, one among several objective vectors Finland, [email protected] provided by the method, which are calculated using elements of the Managers of larger buildings are confronted representation. Then the method provides to with complex investment decisions the DM information about the achievable concerning possible extensions of the energy region (a box) in the objective space system, like photo voltaics, stationary containing solutions in the representation batteries or heat storage. They have to dominating the selected objective vector. In consider a multitude of objectives, for the following iteration, only solutions from example, investment and annual operation this box are used for selecting the direction costs, CO2 emissions and module lifetime. A of improvement. F-NAUTILUS utilizes the computer-assisted optimization and decision same idea but in addition to that, interacts making process promises substantial with the heuristic algorithm running in benefits in this complex problem. In this parallel, by using intelligent agents. The research, we provide decision support in closer is the current objective vector to the building energy system management by Pareto optimal points, the smaller is the applying interactive multiobjective achievable region. By setting the focus of the optimization methods. We consider five population-based heuristic to this box, the objectives (initial investment cost, running method increases the accuracy and density cost, CO2 emissions, resilience to power of the representation manyfold. The quality outages and battery lifetime) for a system of the Pareto optimal set representation upgrade with both hardware additions and significantly improves towards the end of the modifications of system controllers. We use solution process seamlessly for the DM. We a building simulation software based on the present the hybridization scheme in detail as Modelica standard to simulate energy flows well as an implementation of F-NAUTILUS in a facility with different investment options. combined with a multiobjective evolutionary For optimization, we apply evolutionary algorithm, and show the results of algorithms since, in the past, different experiments demonstrating the advantage of evolutionary algorithms have been our method over E-NAUTILUS. This research successfully tested to find viable investment is a part of developing an open source plans for the problem considered. A major software framework DESDEO drawback of evolutionary algorithms is the (desdeo.it.jyu.fi) for interactive multiobjective long computing times (days to weeks) of a optimization methods. single optimization run, which stretches the patience of (high-level) decision makers (DMs) to the limit. In order to speed up the computation, we apply surrogate models. In particular, we focus on Kriging as surrogate models (also known as Gaussian processes) due to their ability to provide uncertainty information about the surrogates. With five

102 conflicting objectives, it is not easy to get a generated that reflect the preferences of the good representation of Pareto optimal decision maker. Many interactive methods solutions and, hence, we focus on applying have been proposed in the literature and interactive methods, which have not been they differ from each other, for example, in applied to this problem before. The benefits the way the decision maker expresses of interactive approaches are threefold. First, preference information, how information is the computation time is reduced because exchanged between the decision maker and the algorithm will focus on those Pareto the method, what kind of subproblems are optimal solutions that reflect the preference formulated to get solutions based on the information of the DM, not all Pareto optimal preference information available and what is solutions. Second, the DM can direct the the stopping criterion. We often say that we solution process to focus on those solutions can stop the interactive solution process that are interesting and does not need to when the decision maker is satisfied and spend time with uninteresting solutions. confident with the final solution but what Finally, and most importantly, thanks to the does this mean? Besides, comparing iterative nature of the methods, the DM can interactive methods is not simple because learn about the nature of the problem and the decision maker plays an important role trade-offs involved as well as the feasibility of and learns during the solution process. Thus, one's preferences. This will increase the the order in which different methods are confidence of the DM in the results of the applied affects the results. To compensate optimization process and increase the this, one would need a large number of chances of actually realizing the final decision makers to use the methods in solution identified. We demonstrate how the different orders. However, the decision problem considered can be solved with an maker must have appropriate domain interactive multiobjective optimization expertise and be somehow responsible for method and discuss the findings. To be more the final solutions. In most real problems, it specific, we apply variants of the reference is not possible to have such a large number vector guided evolutionary algorithm (RVEA), of decision makers. Naturally, students can incorporate surrogates in the consideration, act as decision makers but only if the introduce a new surrogate-assisted problems to be solved have been formulated interactive method and apply it in the so that the students genuinely feel the building energy system design problem. responsibility for the final solution. If we Finally, we compare the quality of the consider non ad hoc methods, we can in solutions obtained to those of a non- principle replace the decision maker by a interactive approach. utility or value function in the comparison. However, this does not necessarily represent 4. How to Compare Interactive all properties relevant to human behaviour Multiobjective Optimization like anchoring, cognitive biases or the need Methods? to change the preferences thanks to learning. On the other hand, ad hoc methods *Kaisa Miettinen, University of cannot be compared even if a utility or value Jyvaskyla, Finland, function was available. To avoid the need of [email protected] having large numbers of decision makers, Francisco Ruiz, University of Málaga, artificial or machine decision makers have Spain, [email protected] been introduced recently for comparing Bekir Afsar, Independent Researcher, interactive methods. However, they do not Turkey, [email protected] yet capture all relevant elements and we need to develop new performance metrics to Interactive methods have proven to be viable be applied with them. In this research, we approaches to solve many kinds of discuss how to quantify the performance of multiobjective optimization problems. This is interactive methods. We consider learning because they enable the decision maker to and decision phases of interactive solution learn about the trade-offs involved, what kind processes separately, since they have of solutions are available and how feasible different objectives. In the learning phase, the preferences are. Furthermore, they can the decision maker should explore and then enhance computational efficiency since only find a region of interest, and in the decision such Pareto optimal solutions need to be phase converge fast in the region of interest

103 identified. A fundamental question is how to same task. It has become very popular in characterize desirable features of interactive practice and it is a hot research topic as well. methods and solution processes. Once we Although Charnes and Cooper have played a have formulated this, we can formulate significant role in the development of DEA indicators to measure the performance and and MOLP, researchers in these two camps can use them in comparing methods. have generally not paid much attention to Aspects to be considered include how well research performed in the other camp. are the preferences reflected in the solutions Neither Charnes nor Cooper attempted to tie generated, how many iterations with the the two fields together. Even Cooper has said decision maker are needed, can we that MOLP and DEA are fully different guarantee Pareto optimality, ability to analyses. Despite many apparent consider different parts of the Pareto optimal similarities, DEA and MOLP (MCDM) research set, confidence in the final solution, insight has developed separately. It is a pity, gained during the solution process etc. This because —despite differences in research is a part of developing an open terminology—DEA and MOLP address similar source software framework DESDEO problems and are structurally very close to (desdeo.it.jyu.fi) for interactive multiobjective each other. In both models, technically optimization methods. speaking, the purpose is to identify efficient points in a certain space and suggest TUE-4-D projections of inefficient points on the basis Tutorial: MOLP vs. DEA - Relatives or of such information. DEA and MOLP should Friends not be seen as substitutes, but rather as complements. In this tutorial, we show that Tuesday 16:00-17:40 - Room: Topkapı the areas DEA and MOLP can cross-fertilize Palace each other. Chair: Özgür Kabak TUE-4-E MOLP vs. DEA - Relatives or Friends Contributed Session: Advanced *Pekka Korhonen, Aalto University, Optimization Techniques and Multiple Finland, [email protected] Objectives Abraham Charnes and William Cooper have Tuesday 16:00-17:40 - Room: Hagia Sophia had a significant impact on the development Chair: Ignacy Kaliszewski of Multiple Objective Linear Programming (MOLP) and Data Envelopment Analysis 1. Different Types of Solutions in (DEA). The work of Cooper and Charnes on Multiobjective Bilevel Programming: goal programming in the late 1950s can be From Optimistic to Pessimistic regarded as pioneers of the MOLP-research Perspectives area. Charnes et al.'s Management Science article from 1955 constituted the first goal *Maria João Alves, University of programming model. The name “goal Coimbra, Portugal, [email protected] programming" was first introduced in Carlos Henggeler Antunes, University of Charnes and Cooper's book: Management Coimbra, Portugal, [email protected] Models and Industrial Applications of Linear Programming, in 1961. It was many years a Bilevel optimization problems model dominating way to solve Multiple Objective hierarchical non-cooperative decision Linear Programming (MOLP) problems. On processes in which the upper level decision the other hand, Charnes and Cooper are also maker (the leader) and the lower level the founders of the Data Envelopment decision maker (the follower) control Analysis. Together with Rhodes, they different sets of variables and have their own published the first article on Data objective functions subject to Envelopment Analysis (DEA) in the European interdependent constraints. The lower level Journal of Operational Research 1978. DEA problem is embedded in the constraints of is characterized as a method for evaluating the upper level problem. Decisions are made the relative efficiency of Decision Making sequentially, as the leader makes his Units (DMUs) performing essentially the decisions first by selecting values for his variables. The follower then reacts by 104 optimizing his objective function(s) on the about possible outcomes. For this purpose, feasible choices restricted by the leader’s new concepts of solutions are proposed to decisions. Thus, the leader needs to consider the SVBP and to the MOBP. Extreme the follower’s reaction to the setting of his solutions (optimistic, deceiving, pessimistic variables since this influences feasibility and and rewarding solutions) and a moderate the leader’s objective function(s) value(s). solution, resulting from the risk the leader is Sequential decision-making processes that willing to accept, are defined for SVBP. can be modeled by bilevel optimization Optimistic and pessimistic Pareto fronts are problems arise in many aspects of resource defined for MOBP. These novel concepts will planning, management and policy-making, be graphically illustrated emphasizing namely the design of pricing policies. A possible pitfalls associated with the multiobjective bilevel problem (MOBP) may computation of those solutions. have multiple objective functions at one or both levels. A special case of MOBP is the 2. Geometry of the Domination semivectorial bilevel problem (SVBP), in Space of a Non-Convex Multicriteria which there is a single objective function at Problem the upper level and multiple objectives at the lower level. The existence of multiple *Andrzej Skulimowski, AGH University objective functions at the lower level problem of Science and Technology, Poland, adds further challenges and difficulties to a [email protected] bilevel problem because the leader has to By a dominating point in the multicriteria deal with the uncertainty related to the optimization problem (F:U->E)->min(Q), follower’s reaction. For each leader’s where Q is a closed, convex, and pointed decision, the follower has a set of efficient cone in the criteria space E, we will mean any solutions. If the leader has no (or has little) point x in E such that x+Q intersects the set knowledge about the follower’s preferences, FP(U,Q) - the set of nondominated elements it may be very difficult for him to anticipate in the above problem. The set of dominating the follower’s choice among his efficient set. points will be denoted D(U,F,Q). While the Thus, the leader can adopt a more optimistic geometry of FP(U,Q) and of the Pareto set or more pessimistic attitude taking into P(U,Q) in U has been studied by many account his anticipation of the follower’s authors, the properties of D(U,F,Q) are far decision, which can be more or less less known and focus on some subsets of favorable to the leader’s interests. An D(U,F,Q) only. The best known subset of optimistic attitude assumes that the D(U,F,Q) is the set of totally dominating follower’s choice (among his efficient set points T(U,F,Q), by definition x is an element restricted by the leader) is the “best” for the of TD(U,F,Q) iff FP(U,Q) is contained in x+Q. leader, while a pessimistic perspective When used in distance scalarization or assumes that the follower’s choice is the compromise programming, under some well- “worst” for the leader. In SVBP the best known conditions imposed on the distance (worst) solution for the leader is the one that and Q, elements of TD(U,F,Q) serving as presents the best (worst) value for the upper reference points guarantee that the level objective function. In MOBP with distance-minimizing solution is multiple objective functions at both levels, nondominated. The ideal points in E are the concept of pessimism has not been defined (by most authors) as (-Q)-optimal clearly defined in the literature. Actually, the points of TD(U,F,Q). It is far less known that works reported in the scientific literature so-defined ideal points need not be unique. devoted to SVBP and MOBP have mainly The first part of the paper will be devoted to addressed the optimistic approach. the presentation of the ideal point However, it is seldom realistic to consider properties. We will also define the local ideal that the follower is indifferent to all efficient points, which constitute another important solutions resulting from a given leader’s subclass of D(U,F,Q) for non-convex F(U) and decision or that the follower chooses provide the properties. We will also recall the according to the leader’s objectives. The uniqueness conditions for the ideal point and combination of different leader’s attitudes apply it to study the geometry of local ideal and follower’s responses results in different points. Furthermore, we provide new types of solutions that should be exploited to properties of generalized ideal points (GIP), provide useful information to the leader

105 i.e. subsets of E defined as Cartesian money) has run out. If this happen, in general products of nondominated sets with respect (with exception for special cases for which to a minimal covering of the set of criteria primal-dual optimization schemes apply) no {F1,...,Fn}, G={G1,...,Gm}, with information is available how close is the best Gi={Fi(1),..,Fik(i)}. F=(F1,...Fn) refers to the feasible solution derived thus far to the above formulated multicriteria problem, optimal solution. In mulitiobjective where E is assumed finite-dimensional. A optimization the situation is the same -- in special attention will be paid to the geometry general no information is available how close of proper GIPs, i.e. GIPs such that the are the nondominated feasible solutions covering G is disjoint (or, in other terms, G is derived thus far to the efficient solution set. a partition of the set of criteria). We will In this work, we show how to provide such provide the connectedness conditions for information in the multiobjective context by such GIPs and show visualizations of some exploiting the classic notion of relaxation. To characteristic proper GIPs in 4D. Then, in the this aim we make use the concept of lower second part of the paper, we will study the shells and upper shells, developed in our properties of other salient subsets of earlier works. We show that a number of D(U,F,Q), namely the sets of partly optimization problems, stemming from (PD(U,F,Q):=D(U,F,Q)\TD(U,F,Q)) and strictly navigating over the Pareto front, can dominating points, SD(U,F,Q). The latter is mutually ''cooperate'' in populating upper defined as follows: SD(U,F,Q) is the set of shells leading to progressively tightening elements x of E such that (x+Q) intersected bounds (lower and upper) on objective with FP(U,Q) equals P(FP(U,Q) intersected functions. No specific assumptions about the with (x+Q), Q). This set is important because problem to be solved are made. We any strictly dominating point yields a illustrate the proposed approach on nondominated solution of a distance multidimensional knapsack problems minimization procedure, under similar derived from singleobjective knapsack conditions concerning the distance in E and problems from the Beasley OR Library with Q as in case of ideal points. Finally, we sizes up to 30 constrains, 500 variables and provide some properties concerning the 100% dense constraint matrices. We mutual situation of GIPs and the strictly address cases when a top-class commercial dominating points which allow us to derive Q- mixed-integer linear solver fails to provide minimality conditions of distance Pareto optimal solutions. minimization procedures with respect to proper GIPs. These results may also be 4. An Efficient Procedure Interior applied to provide Pareto optimality Point Method Based on a New conditions for the equilibria of mixed Kernel Function for Linear cooperative games, where several disjoint Complementarity Problem coalitions seek consensus solutions. Another application of GIPs is the detection of *El Amir Djeffal, University of Batna, redundant criteria in a multicriteria problem, Algeria, [email protected] which can be related to the geometric properties of the domination space. In this paper, we present an interior point algorithm for solving an optimization 3. Cooperative Multiobjective problem using the central path method. By Optimization with Bounds on an equivalent reformulation of the central Objective Functions path, we obtain a new search direction which targets at a small neighborhood of the Ignacy Kaliszewski, Polish Academy of central path. For a full-Newton step interior- Sciences, Poland, point algorithm based on this search [email protected] direction, the complexity bound of the *Janusz Miroforidis, Polish Academy of algorithm is the best known for linear complementarity problem. For its numerical Sciences, Poland, tests some strategies are used and indicate [email protected] that the algorithm is efficient An optimization problem is numerically not solvable if the optimal solution has not been derived because the budget (time, memory,

106

TUE-4-F the total radar detection threat of the UAV fleet. In the two-objective version, there is a Invited Session: Multiobjective Route single efficient trajectory between target Planning pairs. In the three-objective case, there are infinitely many trajectories connecting target Tuesday 16:00-17:40 - Room: Basilica pairs, each trajectory having a better Cistern distance or radar detection threat value Chair: Diclehan T. Öztürk, Murat Köksalan compared to another. For the two-objective version, the aim is to find the visiting order. 1. A Heuristic Approach to Multi- On the other hand, for the three-objective Objective Routing Problem for a case, we need to determine both the visiting Fleet of Unmanned Aerial Vehicles order of the targets and the trajectories to use between target pairs. For both problems, *Büşra Bişkin, Gazi University, Turkey, we aim to find all nondominated solutions. [email protected] We employ an evolutionary algorithm, a Diclehan Tezcaner Öztürk, Hacettepe modified version of NSGA-II, to find an University, Turkey, approximation of the nondominated frontier. [email protected] We develop specialized representation, Ceren Tuncer Şakar, Hacettepe crossover, and mutation operators in NSGA- University, Turkey, II to address our problem. Each time a new [email protected] solution is generated, we enhance it using a heuristic method. We develop a repair In this study, we aim to find routes for operator for solutions that violate the multiple Unmanned Aerial Vehicles (UAVs). constraints. To handle the infinitely many The UAVs start from a common base, visit a trajectory options between target pairs, we number of targets and return to the base in develop some simplifications for the terrain a two-dimensional terrain. Each UAV can representation. We also consider utilizing the travel a limited distance since there is a fuel preferences of the decision maker in the limit on UAVs. Nowadays, UAVs are used for search process by modifying the crowding several missions. At each mission, different distance measure in NSGA-II with preference objectives and problem structures should be values of solutions. considered. As an example, in military operations like espionage and 2. Interactive Algorithms for reconnaissance, exposure of UAVs to enemy Biobjective UAV Route Planning threats like radars are considered an Problem in Continuous Terrain important risk factor. Therefore, not only the number of visited targets should be *Hannan Tureci, Middle East Technical optimized but also the duration of the threat University, Turkey, exposure should be considered. In [email protected] emergency aid operations, the most Diclehan Tezcaner Öztürk, Hacettepe important objective is arriving at the targets University, Turkey, as fast as possible. Especially in these [email protected] missions, the UAVs should arrive at their final Murat Köksalan, Middle East Technical destination in the shortest time. Since in University, Turkey, some operations the targets do not have the same importance, priority to the targets [email protected] should be given. In our approach, we We study the unmanned air vehicle (UAV) consider the following objectives; minimizing route planning problem in the continuous the total distance traveled by the UAVs, terrain, where the UAV is allowed to move to minimizing the radar detection threat of the any point in the terrain. We minimize the total UAVs, and maximizing the total reward distance traveled and the total radar collected from the targets visited. We first detection threat. Two objectives and consider the two-objective version of this movement in continuous terrain leads to problem where we maximize the total reward infinitely many trajectory options between collected from the targets while minimizing target pairs. The overall problem has a the total distance traveled. In the three- continuous nondominated frontier. We objective version, we additionally minimize develop interactive algorithms that find the

107 most preferred solutions of a decision maker are located in a continuous terrain. A UAV (DM) for the cases of linear and quasiconvex starts its movement from a base, visits all underlying preference functions. In the targets and returns to the initial point. The underlying linear preference function case, terrain is monitored by radars. The aim in this we search for supported efficient solutions. problem is to determine both the visiting The idea of the algorithm is to ask for order of the targets and the specific pairwise comparisons and to update the trajectories to be used between consecutive weight space based on the DM’s responses. target pairs under multiple objectives. We We start with two weights that divide the considered the bi-objective and the three- weight range into three equal-length objective versions of the problem. In the bi- intervals and find the corresponding tours. objective version, we minimize both the We update the weight range based on the distance traveled and the radar detection DM’s choice between the tours and continue threat. We included maximization of in a similar fashion. At the end, we present “satisfaction” as the third objective for the the solution corresponding to the weight that three-objective version. This objective divides the final weight range into two equal- indicates the satisfaction degree of a length intervals as the final solution. In the decision maker (DM) with the quality of quasiconvex case, we conduct our search in information collected (e.g., visual images) two stages. In the first stage, we define from a target when it is visited. The quality of weights dividing the initial weight range into information collected from a target varies n-1 equal length intervals and find the during a 24-hour time period depending on corresponding supported efficient solutions. the weather and light conditions. Such a Similar to the linear case, we ask for pairwise structure is referred as soft time windows in comparisons from the DM and we reduce the the literature. The continuous terrain where search region in the objective space into there are infinitely many trajectory options rectangular regions around at most three for each visiting order to the targets and supported nondominated solutions. In the conflicting objectives lead to infinitely many second stage, we search inside these Pareto-optimal solutions. Generating all rectangular regions. We use mathematical those solutions is neither computationally models that exclude the inferior regions to practical nor meaningful. We develop a obtain new candidate nondominated preference-based multi-objective solutions. We terminate both algorithms evolutionary algorithm (MOEA) that converge when the DM finds the two presented to the preferred regions of the Pareto-optimal nondominated solutions too close to choose frontier. Preferences of a DM are elicited one of them. We demonstrate both through reference points in the objective approaches on different UAV route planning space and the algorithm converges to problems with 5 and 9 targets. The results regions of the Pareto-optimal frontier close to show fast convergence to the preferred the reference point(s). Reference point- regions in few pairwise comparisons. based MOEAs typically depend on the initial preferences of the DM, and are sensitive to 3. A Preference-based Multi- the properties of the reference point Objective Evolutionary Algorithm to (nondominated or dominated) and the Bi and Three-Objective UAV Route structure of the Pareto-optimal frontier Planning Problems in Continuous (continuous or discontinuous). Our algorithm Space provides flexibility in terms of reference point definition. The performance of the algorithm *Erdi Dasdemir, Hacettepe University, is robust to the type of the reference point Turkey, [email protected] and the structure of the Pareto-optimal Diclehan Tezcaner Öztürk, Hacettepe frontier. Multiple reference points may be University, Turkey, defined simultaneously and the algorithm finds close solutions to each reference point. [email protected] The DM can also change his/her reference Murat Köksalan, Middle East Technical point during the algorithm and continue the University, Turkey, search in different regions. Developed [email protected] archive mechanism implicitly stores a We address a military-type UAV route representative set of nondominated planning problem where targets and radars solutions found throughout the algorithm. 108

This set can be presented to DM whenever When the UAV is at the base, we solve a he/she wishes. The information on the traveling salesperson problem to find a spread of the Pareto-optimal points provides route. The problem then reduces to a insights to the DM for exploring different shortest Hamiltonian path problem when the regions by changing the reference points. In UAV is at one of the targets. The preparation addition to the preferred solutions of the DM, phase before solving the new problem the algorithm is also able to present a set of requires considerable computational effort. solutions representing the whole Pareto- To obtain a good solution reducing the optimal frontier at the end. Addressing the computational burden, we develop k-closest UAV route planning in a continuous terrain heuristic. We demonstrate the heuristic on also required the development of new two UAV route planning problems with 5 and mechanisms for solution representation and 9 targets. We obtain good results in short fitness assignment. To the best of our durations. knowledge, this is the first study that adapts a preference-based MOEA to a UAV route planning problem with multiple targets in continuous terrain. We tested the algorithm on several hypothetical problems. Results show that our algorithm is able to approximate the true Pareto-optimal solutions independent of the shape of the Pareto-optimal frontier and regardless of whether the reference point is dominated or not. We also develop mechanisms specific to the UAV route planning problem and demonstrate the algorithm on several problems. Results show that our algorithm converges to preferred regions on the Pareto- optimal frontier and adapts to changes in the reference points quickly. 4. Real-Time Biobjective UAV Route Planning in Continuous Terrain *Diclehan Tezcaner Öztürk, Hacettepe University, Turkey, [email protected] Murat Köksalan, Middle East Technical University, Turkey, [email protected] Nail Karabay, The University of North Carolina at Chapel Hill, United States, [email protected] We consider the route planning problem of an unmanned air vehicle (UAV) in the continuous terrain. In this problem, the UAV starts from a base, visits all targets, and returns to the base. During the mission period of the UAV, the locations of the targets change. We find the best route of a decision maker (DM) whose underlying preference function is linear, considering two objectives; minimizing distance traveled and minimizing radar detection threat. We develop a real- time algorithm that reconstructs the route of the UAV each time the UAV visits a target.

109

point out new properties of the net flow scores which tend to legitimize or WEDNESDAY delegitimize the use of this method. Finally, a better interpretation of these flow scores WED-1-A will allow understanding better how to use the different preferential parameters used in Contributed Session: Outranking Promethee as well as their influence on its Methods final outcome. Wednesday 9:00-10:40 - Room: Galata 2. Outranking Relation-Based Tower Aggregation and Modelling of Time- Chair: Marc Pirlot Varying Preferences and Data for Multicriteria Decision Making 1. Some Considerations about the Legitimacy of Using Flow Scores to *Salem Chakhar, University of Compute PROMETHEE Rankings Portamouth, United Kingdom, [email protected] *Gilles Dejaegere, Université libre de Anissa Frini, Université du Québec à Bruxelles, Belgium, Rimouski, Canada, [email protected] [email protected] Mohamed Aymen Boujelben, University Alessio Ishizaka, University of of Sfax, Tunisia, Portsmouth, United Kingdom, [email protected] [email protected] Yves De Smet, Université libre de Ashraf Labib, University of Portsmouth, Bruxelles, Belgium, United Kingdom, [email protected] [email protected] Multicriteria decision aid consists in helping Vincent Mousseau, CentraleSupélec, decision makers to compare (rank, choose, France, sort, etc.) different alternatives which are [email protected] evaluated on conflicting criteria.These last decades, numerous decision aid methods A common assumption when addressing a have been developed. They can be divided in multicriteria decision problem is to suppose three main different categories: aggregating, that preferences and data is unchanging outranking and interactive procedures. One over time. This can be justified in some family of the well-known outranking decision situations but not in general. procedures are Promethee methods. These Indeed, in several situations, the dynamic work as follows. First, all pairs of alternatives and evolving nature of real world is are compared on each criterion leading to considered to have no effects on the unicriterion preference values. These are outcomes of the decision-making process. In then aggregated using a weighted sum to practice, however, the high equity of build a valued pairwise preference matrix. It decisions and their long-term impacts on is then exploited to assign a positive, a population, organizations, etc., impose, to negative and a net flow score to each some extent, an explicit incorporation of the alternative. These scores are then used in dynamic nature of real world into problem order to get a partial (PROMETHEE I) or formulation, modelling and resolution. complete (PROMETHEE II) ranking. Despite Accordingly, we can distinguish two different the wide use of these methods in practical perceptions of the decision-making cases, there is a great lack of theoretical environment: (i) a static perception in which works about the nature of net flow scores. the inherent dynamic nature of the decision The aim of this work is to provide some new problem are ignored (or not recognized) insights on the possible interpretations of because they have no significant effects on these net flow scores. This will help the the achievement of the decision-making decision analyst and the decision maker to process, or because their handling is decide whether Promethee methods are well expensive and/or complicate, or (ii) a suited for the decision problem and the dynamic perception in which the evolutionary decision context at hand. In addition, we nature of the decision environment is

110 explicitly integrated in the decision making strategy where we proceed by an aggregation process. Several researchers have been with respect to the criteria dimension interested in the study of time effect on followed by an aggregation with respect to decision-making and different theories have the time dimension; and (iii) mixed been developed, including time preference, aggregation strategy where both dimensions intergenerational time preference, are jointly considered. The proposed intertemporal choice, time series analysis, approaches are illustrated and validated dynamic/real time decision making, using real world data. Results show that the sequential decision-making and stochastic proposed approaches permit to avoid the decision making. These researchers cover a problems encountered by functional large spectrum of domains, including aggregation of time-varying preferences and economy, banking, finance, health, life data and handle most of temporal saving, insurance, business, environment semantics. and climate change. Earliest investigations have been conducted in the economic 3. Beyond Multicriteria Ranking context where attention is focused on the Problems: The Case of PROMETHEE study of the effect of realisation of outcomes *Yves De Smet, Université libre de on the consumer preference. These initial works are based on the definition of a set of Bruxelles, Belgium, axioms that contribute to the construction of [email protected] time-dependent utility functions. The basic PROMETHEE is a well-known multicriteria concept of these studies is actualisation that outranking method that has been applied in postulates that the desirability of outcomes hundreds of applications. Its success is due decreases over time. Later works concern to its simplicity and the existence of user the study of consumer behaviour in dynamic friendly software such as Visual economic environment. The effect of time on PROMETHEE, Smart Picker or D-SIGHT. If it consequences/preferences is also was primarily developed for (complete or addressed in the context of inter- partial) ranking purposes, recent extensions generational choice, where decision often have been proposed in sorting and clustering needs a compromise between current and contexts. Among them, the methods called future generation. Most of the works in this PROMETHEE TRI and PROMETHEE CLUSTER domain are empirical and show that were first presented in 2004. Unfortunately, individuals give more importance to current these approaches suffered from some generation than future ones. This paper drawbacks that we highlight in this addresses the problem of modelling time- contribution. To overcome these problems, varying preferences and data in the context authors have developed other extensions of multicriteria decision making. In dealing such as FlowSort, PCLUST, etc. At first, the with time-varying preferences and data, an purpose of this presentation is to provide a important question arises: how can we summary of these contributions, to highlight represent the temporal preference their existing links and list several remaining semantics induced by the time dimension research questions. Then, from a more and how aggregate them coherently? In general perspective, we will show how a conventional preference modelling, we may ranking method can easily be adapted to be distinguish three basic aggregation applied in a sorting or in a clustering context. strategies: (i) the use of utility/value In addition, we will illustrate how sorting functions, (ii) the use of binary relations, or methods can be used to compute complete (ii) the use of if-then decision rules. In this or partial rankings. These concepts will be paper, we propose a series of approaches to illustrated in the specific case of deal with time-varying data and preferences. PROMETHEE (but are not limited to it). More These approaches are based on extended globally, we will see that the boundaries versions of the outranking relation and adopt between the three fundamental problem different aggregation strategies: (i) criteria- settings (ranking, choosing and sorting) are oriented aggregation strategy where we blurred. Finally, issues related to rank proceed by an aggregation with respect to reversal problems will also be addressed the time dimension followed by an from this new point of view. aggregation with respect to the criteria dimension; (ii) time-oriented aggregation

111

4. Assigning Alternatives to the Good 1. Digital Transformation Roadmap or Bad Category based on Several in Manufacturing Limit Profiles *Onur Öztürk, ÇİMTAŞ, *Marc Pirlot, University of Mons, [email protected] Belgium, [email protected] Digital technologies aimed at manufacturing Denis Bouyssou, CNRS-LAMSADE, processes are proliferating with the Université Paris-Dauphine, France, transformation of conventional business [email protected] models. While multinational technology Thierry Marchant, Ghent University, collaborations set the pace of the process, Belgium, [email protected] investments made in hardware infrastructure and human resources have Recently, Fernandez et al. (EJOR 2017) have become the new foundation for sustainable proposed a method for sorting alternatives growth. For companies which determinedly into ordered categories based on several press on with digital transformation in criteria. This model known as ELECTRE Tri-nB manufacturing, a correct road map and is a variant of ELECTRE Tri. Instead of using leadership are the keys to success, a single limit profile to determine whether an especially during these times which is the alternative reaches at least a certain quality emergence of a new technologic revolution. level, it uses several. In this work, we Çimtaş operates in various sectors ranging consider a simplified version of ELECTRE Tri- from power, oil & gas to aviation. We strive to nB, which lends itself to an axiomatic increase our efficiency and facilitate the lives characterization. Dropping some pecularities of all our stakeholders by integrating digital of the underlying model, as originally technologies into all our processes. Last presented, while keeping its essence permits year we started our journey of digital to shed light on the main features of the transformation by initially evaluating model and to analyze its properties. More previous success stories in various sectors specifically, we characterize the assignments and choosing our technology partners. We to ordered categories that can be described gathered a comprehensive team consisting by the simplified model. This is done in the of people from a wide range of disciplines in spirit of previous work by the present order to develop our digital road map and authors, in particular the characterization of define our priorities. We have undergone a outranking relations and the non- huge technologic transformation, initially compensatory sorting (NCS) model (EJOR about data gathering. In addition by putting 2007). In this talk, we shall: - give a flavor of the tailor-made MES and Condition the axioms used to characterize the model; Monitoring platforms into effect, which - analyze the model complexity in terms of combine digitalization of customer the number of profiles needed to separate processes and operations we have two ordered categories; - tackle algorithmic succeeded to reach out to our entire team. issues such as checking whether a given With the initiation of the second phase of our ordered partition can be represented in the project around 2 months ago, we aim to model and, eventually, eliciting the model's interpret and convert various data collected parameters. from different sources into economic value WED-1-B by prioritizing them via multi-criteria decision-making approach and putting data Special Session: AHP/ANP Applications science into the center of our focus. During in Industry 4.0 - sponsored by ÇİMTAŞ this phase, in addition to our current partners we have included a couple of additional start- Wednesday 9:00-10:40 - Room: Maiden's ups specializing in data science into our Tower ecosystem, all the while increasing our Chair: Nenad Medic international cooperation working with very valuable companies across the globe.

112

2. Strategic Comparison of Additive comparisons were made by experts who Manufacturing and Robotic have experience in production as well as in robotics and 3D printers. Experts made Machining with Analytical Hierarchy comparisons for four different production Process under Different Competition scenarios regarding two different production and Production Scenarios volumes and two different product complexity levels. The study shows that *İrem Özgen, Istanbul Technical different manufacturing technologies have University, Turkey, strategic dominance under different [email protected] scenario groups. Results of the AHP analysis Seçkin Polat, Istanbul Technical indicate that additive manufacturing is University, Turkey, [email protected] clearly dominant in scenarios containing low Umut Asan, Istanbul Technical production volume, regardless of how University, Turkey, [email protected] complex the part is, whereas robotic machining shows its dominance in high This study provides a strategic comparison of volume production of parts with simple additive manufacturing and robotic designs. In the scenarios covering complex machining within the Industry 4.0 context design parts and high volume production, under various competition and production there is no clear dominant manufacturing scenarios. The comparison was carried out technology. using the Analytical Hierarchy Process (AHP). The proposed AHP model helps to determine 3. Applying Fuzzy MCDM Method to under what scenarios which production Select Internet of Things Platform for technology leads to a greater competitive Iranian Companies advantage. The comparison results could be very helpful for companies that plan to use *Masoud Shayganmehr, Tarbiat 3D printers and/or robotics as their Modares University, Iran, production resources, since these resources [email protected] require huge investments. Although there Gholamali Montazer, Tarbiat Modares are studies in the literature comparing either University, Iran, robotic machining alternatives or different [email protected] 3D technologies, there is no study comparing these two manufacturing technologies. Thus, With the advent of Internet of Things (IoT), to the best of our knowledge, this study is the many Industries have decided to utilize this first in this regard. In addition, the study technology in order to increase their offers an analytic modeling and efficiency and performance over the past measurement process to operationalize the years. Various applications of Internet of contribution of these technologies and Things are used in different industries such allows for a comprehensive comparison. In as agriculture, transportation and health. IoT this respect, the criteria set and the is made of three layers including sensor, comparison context is also novel. The communication network and platform. proposed AHP model includes two main Sensors are responsible for accumulating criteria, cost reduction and value creation, data from environment and send it to that are adopted from strategic management platform through communication network. theory. Cost reduction consists of the Platform is the closest layer to end-user following sub-criteria: material, labor force, which is responsible for data analysis and energy, extra tooling and maintenance costs. delivering services to different industries and Design, quality, support, image and delivery users. One of the most important managerial are the sub-criteria of value generation. The task is selecting an appropriate platform alternatives compared in the AHP model are based on the following criteria: "number of robotic machining and additive supported protocols", "security, maintenance manufacturing. No comparison was made and after sales services", "integration with between cost and value creation or between other IoT platforms", "scalability" and "price" their sub-criteria by experts. Importance because there are numerous IoT platform values of the main and sub-criteria was given vendors and astronomically high primary as scenarios that are called competition price of each one make the decision much scenarios, by the authors. All other more difficult. In this research, a multi

113 criteria decision-making method is applied to which often involves single units of complex solve the problem in order to mitigate the products made according to customers’ complexity of decision-making. Additionally, orders. Concept of Industry 4.0 is considered due to the subjective proposed criteria, Fuzzy as an enabler of these trends in production. Set Theory and Grey Numbers are utilized for In order to adopt to these trends in most modeling qualitative comments of experts. efficient and profitable way, companies The research has been conducted for one of should be able to recognize organisational Iranian Company called "Faraz" which and technological concepts that contribute intended to use one of top-leading IoT the most to this kind of production. Also, this platforms to develop their own applications. should be done in a way that does not The ultimate result of this research indicates disregard organisational and technological that the most important criteria in IoT concepts that are not oriented towards platform selection are "security", "price", aforementioned trends in production, as this "number of supported protocols", "ability of is not the only possible way to meet integration with other platforms", "scalability" customers’ needs. The focus of this research and "maintenance and after the sales is to develop a set of criteria that could be services" respectively and Microsoft platform used for evaluation of organisational and was selected as the most suitable platform technological concepts based on product for Iran’s Industry among top-leading IoT and production characteristics. Those platform vendors in the world including characteristics include product development "Oracle", "Microsoft" and "Amazon". strategy, manufacturing strategy, level of product complexity, and batch size. In this 4. Determining Importance of research, they are considered as 4 Production Characteristics Using dimensions of the imposed problem. Fuzzy AHP: Industry 4.0 Perspective Subsequently, these dimensions are elaborated in more details in order to *Nenad Medic, University of Novi Sad, encompass all relevant criteria for evaluation Serbia, [email protected] of organisational and technological Bojan Lalic, University of Novi Sad, concepts. As a result, 12 criteria are Serbia, [email protected] identified, namely: product development Zoran Anisic, University of Novi Sad, according to customers’ specification, Serbia, [email protected] product development as a standardized Nemanja Tasic, University of Novi Sad, basic program into which customer specific Serbia, [email protected] options are implemented, product development for a standard program from Ugljesa Marjanovic, University of Novi which customer can select, manufacturing Sad, Serbia, [email protected] upon receipt of customer’s order (i.e. made- Industry 4.0 has gained a lot of interest in to-order), assembly upon receipt of recent years. Most of the research related to customers’ order (i.e. assembly-to-order), this topic is oriented towards implementation manufacturing to stock, simple products, and use of emerging technologies in products with medium complexity, complex manufacturing in order to make processes products, single batch, small batch, and digitized and automated. Indeed, new large batch. Furthermore, this research aims technologies bring changes in manufacturing to determine the importance of these criteria environment. However, it was determined by in the context of Industry 4.0. For this vast number of researchers that there are purpose, Analytic Hierarchy Process (AHP) some prerequisites for appropriate was used. This method is considered as the implementation of advanced technologies in most appropriate from the perspective of production processes. Organisational hierarchically structured problem, as it is the changes (i.e. use of innovative organisational case in this research. Also, fuzzy logic was concepts) have been recognized as an introduced in the evaluation process in order important factor for manufacturers to adopt to reduce the vagueness and uncertainty of to emerging trends of production. In constant decision makers’ judgment. In this research struggle to maintain their competitiveness we used triangular fuzzy numbers, since they on the market, companies are focused on are the most utilized in fuzzy multiple satisfying customers’ needs. For this reason, attribute decision making studies, due to companies engage in custom production their computational simplicity and suitability

114 to the nature of experts’ linguistic methods for environmental risk evaluation, evaluations. Experts from manufacturing none of the methods considers human risk companies were involved in the evaluation attitudes, and decision-making process of process. The results indicate that in the human is generally accepted as rational, context of Industry 4.0, criteria with highest such as the utility theory based perspective. complexity are considered as the most Therefore, human decision-making process important. These results show that under risk and irrationality is not usually manufacturers are aware of the fact that new taken into account, and decisions are trends in production represent an important considered to depend only on the outcome. factor for achieving sustainability, However, in real life, decisions are not always development, and increased rational, and the decision makers can be competitiveness of their companies. Results risk-neutral, risk-averse or a risk-seeker. presented in this research are particularly Therefore, to account for irrationality and risk important for strategic orientation of attitudes of decision makers under risky and manufacturers. Moreover, these results uncertain environment, Prospect Theory is could be used for further research that proposed in this study, to create sustainable involves evaluation and analysis of the use of risk assessment. After theoretical organisational and technological concepts in explanations, appropriate methodology is manufacturing companies. needed to evaluate risks of e-waste recycling. For decision making under risk, WED-1-C prospect theory based multi criteria decision- Contributed Session: MCDM for making method (MCDM), TODIM, is Environmental Decisions recommended in this study. The word TODIM is an Portuguese acronym in for Interactive Wednesday 9:00-10:40 - Room: and Multi Criteria Decision Making. Unlike Dolmabahçe Palace most of the multi criteria methods, which assume that decision makers seek the Chair: Pınar Darende maximum global measure value for the 1. Sustainable E-Waste Recycling corresponding solution, TODIM utilizes global from Risk Assessment Perspective measurement of value calculated by the application of Prospect Theory. TODIM gives with TODIM the shape the value function similar to Yeşim Deniz Özkan Özen, Yaşar gains/losses function of Cumulative University, Turkey, Prospect Theory, where gains and losses are [email protected] always recognized concerning a reference point. Because it represents the limited *Yiğit Kazançoğlu, Yaşar University, rationality of the decision maker during the Turkey, [email protected] decision-making process, TODIM is a Sachin Kumar Mangla, Plymouth beneficial method for behavioural decision- Business School, United Kingdom, making. The main approach of the TODIM is [email protected] to define the relative degree of dominance for each alternative by using prospect theory Electronic waste is a growing problem due to based utility function. This study involved a shortening product life and increasing rates numerical example, a computer disassembly of consumption and technological problem. Aim of the problem is to find the development. Therefore, due to optimal sequence of the disassembly tasks, environmental concerns, recycling of which indicate parts of computer. In the electronic products is attracting more numerical example there are 12 risks, which attention. E-waste contains many different are named as criteria, and they are used to hazardous materials and e-waste recycling evaluate 9 tasks of computer disassembly activities have risks for human and processes, which are taken as alternatives. environmental safety, and businesses. Aim At the end of the implementation, tasks are of this study is to focus on sustainable e- prioritized by considering risk factors. To sum waste recycling from risk assessment up, first novel point of, this study categorizes perspective by considering decision-makers’ the risks of e-waste recycling as risk attitudes and irrationality under risk and environment, human and business. The uncertainty. Although there are different second novel point is to propose Prospect

115

Theory to consider irrationality, and risk soil conservation in Oman, decision-making attitudes of decision makers within the risky governmental departments and research and uncertain nature of e-waste recycling. institutes. The expert advisory group was Related to the second novel point, Prospect formed based on the related knowledge Theory based TODIM method is proposed for background and experience in the field of risk evaluation in e-waste recycling. From the study, which involved specialists from managerial implications point of view, e- different organizations. The group comprises waste recycling is an economically promising managers with duties of performance sector however, inherent potential risks monitoring and measurement should be considered by organizations in responsibilities, experts from operational order to achieve sustainability by preventing levels such as engineers and technicians, environmental, human and business losses. academic staff, and consultants. The hierarchy is arranged in such a way that it 2. Application of Analytical Hierarchy has three levels. The top level consisted of Process for Management of Soil the goals of the current study. The middle Erosion in Oman level comprised main criteria including the types and characteristics of soil in Oman Alia Al-Rahbia, Middle East College, (rocky and sandy), the climate of Oman Oman, [email protected] (arid/semi-arid climate) and the cost of the Mohammed Abushammala, Middle East soil protection methods. Meanwhile, the College, Oman, bottom level represents alternatives which [email protected] involves stone gabions, land husbandry, *Wajeeha Qazi, Middle East College, windbreaks, hollows or pits, soil conditioning, Oman, [email protected] field cropping practices, erosion control fences, treating footpaths, protection of the The level of soil degradation worldwide is gully head, reshaping donga systems, alarming due to its potential for causing revetments, compost blanket, masonry walls serious problems even to the extent of and brush layering and brush barrier. The threatening food security. Sultanate of Oman results of AHP analysis indicated that the also faces soil erosion which disturbs most suitable soil erosion protection method, waterways, infrastructure, and agriculture. based on the considered factors, is erosion The problem of soil erosion has become control fences followed by protection of the increasingly apparent in Oman due to gully head and revetment practice. The study inappropriate land management, especially also indicated that the AHP model is a in the case of dams where sedimentation is suitable guiding framework for decision a common issue. Currently, there are no making because it involves a group of appropriate practices followed to overcome experts for the selection of soil protection the problem of soil erosion in Oman and no methods based on the situation of Oman. specific studies are available related to the suitable methods for soil protection in Oman. 3. Developing Environmental Policy Hence, this study is an initiative to overcome Instruments for Enhancing Females’ the problem of soil erosion in Oman by Concerns based on the DANP-MV proposing the best soil erosion protection Method method using Analytical Hierarchy Process (AHP). In order to achieve this goal, this study Chia-Lee Yang, National Center for High- reviews different methods of preventing soil Performance Computing, Taiwan, erosion and the factors governing their [email protected] selection. The important and controlling *Chi-Yo Huang, Taiwan Normal factors were considered in an AHP model to University, Taiwan, rank the soil erosion protection methods. In the AHP method, a group of experts was [email protected] formed to identify/modify the soil erosion Ching-Chun Hsu, National Ilan protection methods and the criteria University, Taiwan, [email protected] influencing their selection. After a thorough Yu-Tai Wang, National Applied Research literature review, a consultation process was Laboratories, Taiwan, conducted with 15 experts from different [email protected] sectors; including authorities responsible for

116

In the past decade, air pollution has become The need for energy resources increases one of the most concerned risks, which has rapidly depend on industrial developments, already impacted the life of most human technological developments and population beings from the aspects of environment, growth. In this context, the energy needs of health, etc. During the past years, Turkey especially electricity and heating will researchers have recognized the gender eventually increase in the next years since differences toward air pollution issues. Turkey is one of the developing countries. Females are less concerned about the air The countries are enforced to prefer pollution-related issues. However, females renewable energy resources over time since are more influenced by air pollutions. the decrease in fossil fuel resources and the However, very limited works have tried to negative effects of fossil fuels on human derive factors influencing females’ attitude health and the environment. In this study, toward air pollution. Therefore, this research the selection of renewable energy types for aims to derive the factors influencing seven different regions of Turkey will be females’ attitudes toward air pollution. discussed with considering the efficiency of Further, appropriate environmental the renewable resources of the regions. education policy tools will be defined to During the evaluation process; solar, enhance female’s attitude toward air geothermal, wind, biomass and pollution. A hybrid multi-criteria decision- hydroelectricity energy sources are making (MCDM) model, the DEMATEL Based considered with using Intuitionistic Fuzzy Analytic Network Process integrated with the Technique for Order Preference by Similarity Modified Vikor (DANP-MV), is used to derive to Ideal Solution (TOPSIS) method which is the influence relationships of factors by the intended to determine whether will use of DEMATEL. Then, the weights being renewable energy in the related region. associated with the factors will be derived by Investment priorities of selected plants using the DANP. The Vlse Kriterijumska according to the region are determined and Optimizacija Kompromisno Resenje (VIKOR) sorted by using data, criteria and expert will be used to evaluate and select the most opinions. appropriate environmental policy tools for crossing the gender gap and enhance WED-1-D females’ attitude toward air pollutions. An Contributed Session: Evolutionary empirical study based on Taiwan will be used Algorithms, Preferences and to verify the effectiveness of the proposed MCDM methods. The policy instruments will Applications be defined to raise the level of the most Wednesday 9:00-10:40 - Room: Topkapı dominant factors which can influence Palace females’ attitudes from the current status to the aspired level. These factors, influence Chair: Tomohiro Hayashida relationships as well as policy instruments 1. Maintaining Diversity in being defined will be used to inform policy for Preference based Evolutionary enhancing females' attitude toward air quality. The proposed policy instruments can Multiobjective Optimization be used by the governments. *Bhupinder Saini, University of 4. Examination with an Intuitionistic Jyvaskyla, Finland, Fuzzy Method by Region of [email protected] Renewable Energy Investments in Kaisa Miettinen, University of Jyvaskyla, Turkey Finland, [email protected] Jussi Hakanen, University of Jyvaskyla, Babak Rouyendegh, AYBU, Turkey, Finland, [email protected] [email protected] The objectives of multiobjective optimization *Pinar Darende, TOBB ETU, Turkey, problems (MOPs) are often conflicting in [email protected] nature and we have trade-offs among the Tahir Khanıyev, TOBB ETU, Turkey, objectives. Because of this, instead of having [email protected] a single optimal solution, MOPs have a set of optimal solutions forming a Pareto optimal

117 set (in the decision space) or a Pareto front known evolutionary algorithms like RVEA and (in the objective space). Preference based NSGA-III. This research as a part of techniques for solving multiobjective developing open source framework DESDEO optimization problems have gained (desdeo.it.jyu.fi), will increase the appeal and popularity over the years, where the applicability of decomposition based MOEAs preference information of a decision maker so that the DM can better control the solution (DM) is used to find the most preferred process and gain desired solutions. Pareto optimal solution. Interaction during the optimization process allows the DM to 2. Exact and Heuristic Methods to gain insight about the problem and the trade- Solve a Bi-Objective Problem of offs associated with it. Focusing on a desired Sustainable Cultivation region of interest rather than the entire objective space also significantly reduces the *Angelo Filho, UTFPR, Brazil, computational (and sometimes economic) [email protected] cost of optimization. Hence, utilizing the In the current context of sustainability and a preferences of the DM is preferable to using cultivation practice in agriculture that a posteriori techniques which try to minimizes the environmental degradation, approximate the entire Pareto front. Many alternative ways that avoid the intensive use decomposition based multiobjective of chemical products in combating pests and evolutionary algorithms (MOEAs) have been the excessive use of the soil has been deeply created which find solutions of MOPs by studied. These measures proposed, if decomposing the problems into single adopted, can stop an environmental crisis on objective or comparatively simpler the planet and make agriculture products multiobjective optimization problems. This is healthier as well. In this sense, one of the usually done by using a set of uniformly central focuses in the crop production distributed reference points or vectors which discussed lately is the use of measures that guide the solution process. Decomposition aim sustainable and ecological planning, based EAs have proven to be capable of considering the environmental degradation solving MOPs with up to 10-15 objectives that has occurred in recent years. For this and seem quite promising. One way to reason, the planning of agriculture activities, incorporate preference information into among them the Crop Rotation, has been these algorithms is to redistribute the gaining prominence in the studies aimed at reference points in a non-uniform manner, in sustainable cultivation, since it is one of the a way that reflects the DM’s preferences. means of cultivation whose practical However, such techniques lead to a loss of principles enable ecological and productive diversity in the population. This is agriculture. This practice, once conducted by detrimental to the solution process, as the rural farmers, brings several benefits MOEAs, which try to replicate the process of since the control of pests, pathogens and evolution via natural selection, require a weeds are performed biologically. This diverse population to reach optimality. decrease the action of pesticides, which is Additionally, drastic changes in the harmful to the environment and bring preferences of the DM, which may be measures of soil recovery, making it always expected if the DM is trying to navigate the fertile. In this study, a nonlinear (more objective space to learn about the trade-offs, specifically quadratic) bi-objective 0-1 may lead to a loss of the progress made by optimization model is proposed. The aim is to the MOEA. This happens because there are minimize (i) the proliferation of pests among very few, if any, solutions near the new the planted crops and (ii) maximize the profit preferred location in the objective space, as of the planting schedule in a certain planning the algorithm was focusing on the previous horizon, respecting the constraints of non- preferred location. Even smaller changes in overlapping planting, non-consecutive the preferences of the DM may impair the planting of varieties of the same family and progress of the MOEA. In this talk, we of the planning time equal in all plots. To introduce new techniques of incorporating determine some Pareto optimal solutions to DM’s preferences into decomposition based this problem, we attributed weights to the MOEAs, which alleviate the aforementioned objective-functions involved, performing a problems. We compare the results of weighted sum of the objectives. The scalar application of these techniques on well- problems obtained with this approach were 118 solved, initially, by using a quadratic Mustafa Zahıd Gurbuz, Doğuş programming solver. In a second moment, a University, Turkey, linearization strategy of the original model [email protected] was applied and we also used integer linear Peral Toktas Palut, Doğuş University, programming in the resulting formulation. Turkey, [email protected] Difficulties arising with the algorithms that implement the exact methods were emerged Non-dominated sorting genetic algorithm II in these two situations, due to the high (NSGA-II) is a very well known and very widely number of integer variables in this problem used method for solving multi-objective and the high computational complexity of the optimization problems (MOOP). NSGA-II can model. This motivated the authors, in a third easily handle constrained MOOPs and also moment, to propose a genetic algorithm (GA) was used for location allocation problems in especially developed to solve the originated literature. And more than a decade later, scalar problems. This GA contain two NSGA-III is introduced to researchers in the constructive heuristics unprecedentedly field and also being very widedly used developed and aim to give an effective recently. The main difference between these performance to the proposed metaheuristic. two methods is how they deal with diversity This study brings the following innovative in order to have more wide spread Pareto aspects: 1) Development of a 0-1 objective front solutions for covering wider ranges of mathematical programming model for alternatives. There are several studies on sustainable cultivation; 2) A comparative comparing the performances of these two computational study between the quadratic methods on particular type of problems and and linear formulations of the proposed within this study these two methods are model; 3) Development of a GA and two going to be compared regarding their constructive heuristics to determine some performances on a multi-objective multi approximations to the optimal Pareto constrained location allocation problem solutions for the problem mentioned. which was first presented as a disaster waste Computational results have shown that GA is management model in the literature. efficient, in the sense that it obtains feasible solutions to the problem in a CPU time 4. Interactive Multiobjective inferior to the exact methods. In the case in Reinforcement Learning for Multi- which the exact methods end their running Step Optimization Problems before obtaining the global optimum of the problem, the solutions from the GA proposed *Tomohiro Hayashida, Hiroshima are, on average, better. This allows us to University, Japan, assume that the mathematical model, as [email protected] well as the developed algorithm, are Ichiro Nishizaki, Hiroshima University, excellent tools in the decision making in this Japan, [email protected] field. The mathematical model proposed, Shinya Sekizaki, Hiroshima University, which aims to establish a way of sustainable Japan, [email protected] vegetable production compromised with the Hiroyuki Yamamoto, Hiroshima profit of the agriculture producer, as well as the solution methods proposed, are University, Japan, applicable in real cases, helping decision- [email protected] makers in choosing different alternatives of In Robot that autonomously decides action agricultural production on their farms and based on surrounding environmental contributing to the advance of knowledge in information such as disaster rescue robot the field of Multiobjective Optimization in a and robot cleaner needs to optimize multiple sustainable environment. objectives simultaneously, such as moving 3. A NSGA-II and NSGA-III as fast as possible to the target location, increasing safety, reducing consumption of Comparison for Solving a Location- fuel and batteries, and so forth. However, in Allocation Problem multi-criteria decision making, objectives often conflict with each other, and in such *Kıvanc Onan, Doğuş University, cases, there does not exist perfect optimal Turkey, [email protected] solutions that simultaneously minimize or maximize all the objectives. Instead of a

119 perfect optimal solution, a solution concept, hypervolume-based multi-objective Q- called Pareto optimality, is introduced in learning (HBQL) (van Moffaert et al., 2015) multi-objective optimization, and many and Pareto Q-learning (PQL) (van Moffaert efforts are accumulated to find a set of and Nowe, 2014) to evaluate Pareto optimal Pareto optimal solutions (Zitzler and Thiele, solution set with three indices of 1998). To improve an objective function, one hypervolume, cardinality, and Pareto or more other objectives of Pareto optimal relation, and demonstrated effectiveness of solutions are deteriorated. Moreover, when these methods. The hypervolume (Bader and there exist multiple Pareto optimal solutions, Zitzler, 2009) employed in HBQL is an index a decision maker selects the most preferred for evaluating a Pareto optimal solution set solution. However, a complex procedure is which means the size of a region dominated required to identify the preference structure by obtained Pareto optimal solutions and of the decision maker (Keeney, 1982). If limited by a certain reference point. In this each Pareto optimal solution is interpreted paper, focusing on a property that a value of as a candidate and try to select one solution the hypervolume increases as the number of out of the Pareto optimal solution set, we do Pareto optimal solutions is small in the not necessarily need to identify the neighborhoods, we propose an interactive preference structure of the decision maker. method reflecting the preference of the Then, we can employ an interactive decision decision maker for multi-objective method that derives a so-called preference reinforcement learning in which the solution of the decision maker by using the hypervolume is used for efficiently finding local preference information obtained from diverse Pareto optimal solutions, not for an interactive process with the decision selecting the preferred solution from among maker (Takagi, 2001; Sakawa and Yano, Pareto optimal solution set. 1988). As a solution method for linear or convex optimization problems, it is possible WED-1-E to apply mathematical solution methods Contributed Session: Portfolio finding an exact optimal solution such as the Optimization Models simplex method, the successive quadratic programming, and the generalized reduced Wednesday 9:00-10:40 - Room: Hagia gradient method. However, for non-convex or Sophia discontinuous problems, approximation such Chair: Michalis Doumpos as evolutionary computation is effective. (Higuchi et al., 2006; Sakanashi et al., 1. Portfolio Investments with Mean- 2004). However, it is difficult to apply these Standard Deviation Model and evolutionary computation methods to multi- step optimization problems. For multi-step Moving Pareto Fronts problems, trial-and-error methods using *Przemysław Juszczuk, University of multi-agent systems is effective and Economics, Katowice, Poland, reinforcement learning with bootstrap type [email protected] estimation is often employed (Sutton and Barto, 1989). This paper proposes an Ignacy Kaliszewski, Polish Academy of interactive multi-objective reinforcement Sciences, Poland, learning method for choosing actions based [email protected] on the preference of a decision maker. In Janusz Miroforidis, Polish Academy of previous studies, for applying reinforcement Sciences, Poland, learning to a multi-objective optimization [email protected] problem, after a multi-objective optimization Dmitry Podkopaev, Polish Academy of problem is reformulated into a single- Sciences, Poland, objective optimization problem by using a [email protected] scalarization method with weighting coefficients, single-objective reinforcement We study the decision making problem, learning is employed (Drugan, 2015; where an investor looks for a portfolio of Natarajan and Tadepalli, 2005). However, it given assets aiming at optimizing two is difficult to determine the weight of the conflicting criteria: expected profit and risk of evaluation for each objective beforehand. the investment. The criteria are modeled via Therefore, van Moffaert et al. proposed two objective functions: mean of expected

120 return and standard deviation of the 2. Asset Elimination in Linear- expected return. The considered problem is Quadratic Problems of Financial the well-known Markowitz portfolio selection problem. We present results of our Portfolio Management investigations in the viability of financial Przemysław Juszczuk, University of investing solutions based on the mean- Economics, Katowice, Poland, standard deviation model in the rolling time- [email protected] window regime. For various investing Ignacy Kaliszewski, Polish Academy of scenarios representing different investor’s risk profiles, we experimentally check, Sciences, Poland, whether one can stable returns by investing [email protected] consistently over some period of time as the Janusz Miroforidis, Polish Academy of mean-standard deviation model dictates. Sciences, Poland, Most of the research in the modern portfolio [email protected] theory concerns the static problem, i.e. the *Dmitry Podkopaev, Polish Academy of case where objective function coefficients Sciences, Poland, are calculated once based on time series in [email protected] a given time period. However, analyzing the behavior of the Pareto optimal set over time The bi-objective mean-variance problem of can also be important from the practical portfolio selection formulated and solved by point of view. The lack of such analysis may Harry Markowitz gave rise to a new direction be caused by the deficiency of large, of research connecting financial investing preprocessed sets of data. We try to close with multiobjective optimization. The feasible this gap by creating such a dataset. We have solution set of this problem is a simple, collected a large set of stock price quotation continuous set of n-dimensional vectors data over a wide period of time, including representing shares of n given assets in the most recent quotations. This allowed us to portfolio. The linear and quadratic objective create a large number of instances of mean- functions represent mean and variance standard deviation model rolling time estimates of the portfolio return, windows over a wide time span. We respectively. Several efficient algorithms experimented with the standard version of have been proposed for analytic derivation of the model to remain within a scope of the the set of Pareto optimal solutions of the quadratic programming framework, which original problem. However, the modern offers efficient exact methods to derive practice of finance dictates the need for Pareto optimal portfolios. For each time extending the mean-variance model, window, we created mean-standard bringing to life more complex multiobjective deviation Pareto fronts and their optimization problems. One type of approximations using, respectively, exact extensions is adding more objective and inexact methods. In this manner, we functions describing additional aspects of have built statistics about Pareto front investment. Another type of extension composition and its dynamics over time. In relates to controlling the number of assets particular, we have shown that a specific selected for investment. The modern heuristic, based on pairwise asset/portfolio financial markets include many thousands of combination yields surprisingly good assets making the problem dimension high. approximations of Pareto fronts with On the other hand, the number of assets to portfolios of low asset cardinalities. Besides be invested in can be restricted in each that, we have calculated return from particular case due to various reasons, such investments according to different investor’s as limited cognitive capacities of decision risk profiles. The research is a stepping makers or limited operational capacities of stone towards extended mean-standard financial institutions. Under such settings, deviation models accommodating various the decision maker may need to introduce an side constraints. As such constraints break upper bound on the number of non-zero down the quadratic programming asset shares or other types of constraints framework, approximate Pareto front related to selecting individual assets into the building methods of verifiable portfolio. Introducing such constraints into approximation, like the ones we developed, the model gives rise to mixed-integer became of importance. formulations posing computational

121 challenges in the case of large n. In order to modeling the risk variance. Qi (2017) is address such challenges, we develop perhaps the first to present an analytical techniques for limiting the number of study addressing the multiobjective portfolio selected assets and controlling the selection optimization problem (MOPOP) with more process. In addition to several mechanisms than one quadratic objective function. This of limiting the number of assets known from goes along with Boyd et al. (2016) who the literature, we propose two new recognize the need to model different types approaches which are applicable to of risk measures naturally leading to several multiobjective portfolio problems with any quadratic functions in the MOPOP model. combinations of linear and quadratic However, the authors are not aware of any objectives. One approach relates to study addressing the computation of the introducing a parametric pairwise efficient set for MOPOPs with more than one comparison relation between assets for quadratic objective function. Recognizing eliminating ones which are in a way the importance of convex MOPs, we collect dominated by others. Another approach is a their theoretical properties from the simple iterative heuristic of approximating perspective of parametric convex the Pareto front with a small subset of optimization since scalarized MOPs naturally assets. We propose several strategies of result in parametric single objective combining different approaches in order to programs (par-SOPs). Parametric control the number of assets in the portfolio. optimization involves two groups of We present results of computational unknowns: the optimization variables for experiments demonstrating strength and which the SOP is solved and the parameters weaknesses of those strategies. that represent unknown problem data. An optimal solution and the optimal objective 3. On Convex Multiobjective value to a par-SOP come in the form of Programs with Application to functions of the parameters, and the solution Portfolio Optimization to a par-SOP also includes a partition of the parameter space into invariancy regions for *Margaret Wiecek, Clemson University, which a specific optimal solution function United States, [email protected] and the optimal objective value function are Among multiobjective programs (MOPs) valid. In the context of multiobjective convex problems play a specific role due to optimization, the obtained optimal solutions their elegant mathematical properties, to the par-SOP become (weakly) efficient easier solvability, and relevance to real-life solutions to the MOP and come as functions applications. A major class within convex of the scalarizing parameters, while the MOPs is made up by multiobjective quadratic partition of the parameter space provides a programs (MOQPs) that have come as an full range of the (weakly) efficient solutions extension of well-established and useful for a full range of potential values of the single objective quadratic programs (QPs). scalarizing parameters. We review three well- MOQPs have a special structure that is established scalarization techniques treating amenable to analytical derivations and them as parametric optimization problems algorithmic developments, but more and emphasize their applicability to (strictly) importantly, to mathematical and real-life convex MOPs. For a specially structured applications. They model decision making class of convex MOPs, we propose a problems in science, business, and modified hybrid scalarization that combines engineering such as regression analysis, the weighted-sum method with the epsilon- finance, predictive control, and others. In constraint method with the distinction that finance, biobjective portfolio optimization only some objectives are put into the was initiated by Markowitz (1952) who weighted-sum while the remaining objectives proposed to minimize the predicted variance create additional constraints. The modified of portfolio return as a measure of risk and hybrid problem is useful when the objective to maximize the expected value of the functions of the original MOP come in two portfolio return. Since then many groups with different real-life meanings. The researchers have been studying portfolio scalarization allows the decision maker for optimization in a multiobjective setting but in an independent analysis of tradeoffs among all those studies, the portfolio problems the criteria within each group. Recognizing typically include one quadratic function also the wide applicability of convex MOQPs, 122 we review four state-of-the-art exact consideration. Moreover, we analyze the algorithms for computing their efficient deviations between historical (estimated) solutions and choose one of them that suits efficient frontiers, actual out-of-sample our needs best. The two lines of efficient frontiers, and realized out-of-sample investigation, the modified hybrid portfolio results. The results provide insights scalarization with the accompanying tradeoff into the effectiveness and performance of analysis and the best solver for MOQPs, are multi-objective portfolio optimization models, merged to solve the MOPOP with three or as well as the synergies derived from more quadratic objective functions, a class combining different risk criteria for portfolio of problems that have not been solved management. before. Numerical examples are included. WED-1-F 4. The Robustness of Multi-Objective Portfolio Optimization Models: An Contributed Session: Practical Applications of MCDM in an Empirical Comparative Analysis Interconnected World Antonis Pavlou, Technical University of Wednesday 9:00-10:40 - Room: Basilica Crete, Greece, Cistern [email protected] *Michalis Doumpos, Technical Chair: Sule Onsel Ekici University of Crete, Greece, 1. Evaluation of Potential Fiber [email protected] Infrastructure Regions for Turk Constantin Zopounidis, Technical Telekom Company University of Crete, Greece, [email protected] *Funda Samanlioglu, Kadir Has University, Turkey, The selection of investment portfolios is a [email protected] major topic in financial decision making, with many portfolio optimization models available Cem Ak, Kadir Has University, Turkey, in the literature. These models extend the [email protected] traditional mean-variance framework using a Fırat Koçoğlu, Kadir Has University, variety of other risk-return measures. Turkey, [email protected] Existing comparative studies of such models Ömer Mücahid Çetinkaya, Kadir Has have adopted a rather restrictive approach University, Turkey, focusing solely on the minimum risk portfolio [email protected] without considering the whole set of efficient Ozan Arslan, Kadir Has University, portfolios, which are also relevant for Turkey, [email protected] investors. This paper, tries to move a step forward, focusing on the performance of the Increased use of internet forces internet whole efficient set, rather of a single service companies to provide innovative portfolio. To this end, we examine the out-of- solutions. In order to meet 'faster internet' sample robustness of efficient portfolios demand of customers, Turk Telekom, the derived by popular optimization models, first integrated telecommunications operator namely the traditional mean-variance model, in Turkey, is planning to expand fiber optic mean-absolute deviation, and conditional infrastructure in Istanbul, Turkey. The value at risk. Moreover, several multi- purpose of this study is to evaluate the objective formulations are considered, which alternatives and select the best potential combine different risk criteria through a fiber infrastructure region in Istanbul. Here, weighted Chebyshev scalarization approach. eight region alternatives in Istanbul are Tests are conducted using data for S&P 500 evaluated based on four criteria which are stocks over the period 2005-2016. The cost efficiency, service quality, estimated results are analyzed through multiple number of customers and infrastructure performance indicators from portfolio and convenience. To evaluate and rank the multi-objective optimization theory to assess alternatives, as the MCDM method, fuzzy the risk-adjusted performance of the AHP is integrated with fuzzy PROMETHEE. portfolios and compare the efficient frontiers Fuzzy AHP is applied to determine the derived from the models under importance weights of criteria and Fuzzy

123

PROMETHEE is implemented to rank the hierarchical structure of the considered region alternatives based on the determined aside the AHP, have created. In an AH weights. Utilizing the integrated Fuzzy AHP- process, the comparison can be made either PROMETHEE approach, the best location by collecting one-to-one customer opinions alternative is found as Eyup and the via the questionnaire or by contacting complete order from best to worst alternative relevant experts. In order for the study to be is determined as Eyup, Umraniye, fair, it was targeted to meet with two experts Basaksehir, Beylikduzu, Sancaktepe, Turkcell and Türk Telekom’s Marketing Esenyurt, Kucukcekmece and Arnavutkoy. department. The interviewers filled the matrices together fairly. The results matrix 2. Analytical Hierarchy Method For was obtained by bringing together the Decision Making In Turkey’s matrices filled by the experts and it was Telecommunications Sector revealed to what extent certain groups of professions were influenced by the criteria in *Bülent Cerit, Istanbul Technical the selection of operators. The results of the University, Turkey, [email protected] study showed that the gender and Kübra Korkut, Istanbul Technical educational status criteria, which were set as University, Turkey, a demographic criteria, did not have a [email protected] distinctive effect on people's choices. Age affects especially the students and non- In recent years, one of the sectors that business owners; gender affects the non- gained the most momentum with the rapid business and public employees. Private development of technology has been the Sector is respectively affected by Telecommunication Sector. Customer Communication Quality, Extra Benefits, preferences are also very important for Quality of Service, Brand Value, Price and companies in the telecommunications Environmental Impact. In general, public sector. Companies attach great importance employees are affected by Communication to the opinions of customers and they want Quality, Extra Benefits, Quality of Service, to respond to their needs in the fastest way Price, Brand Value and Environmental because competition is high, alternative Impact respectively. Self-Employed is operators are rare and innovation is quickly respectively influenced by Communication copied in the sector. One step that can put Quality, Service Quality, Extra Benefits, Price, operators at the forefront of this competitive Brand Value and Environmental Impact. market is to anticipate the criteria by which Although the general characteristics of these customers can give importance to certain three occupational groups are similar to each characteristics for a telecom operator other, it has been observed that the value of selection. Hence, with this information the brand is high due to the fact that the operator can direct the customer to the price criterion is insignificant and the relevant product with a dominant marketing environmental value is high. The fact that the power at the decision stage. The main Quality of Service criteria is of high purpose is to understand how demographic importance in the freelance employee is characteristics and specific criteria that interpreted as the fact that each problem is affect a customer's choice in an operator based on the importance of the customer's selection process. In the first phase of the categories and the continuous contact with project, the decision making methods and the customer service. Non-Business people the criteria affecting the selection of people do not have a decent income structure, price which are Communication Quality, Price, for themselves is an important criterion. For Extra Benefits, Service Quality, Brand Value the students, it was observed that the price and Environmental Impact in the literature was of high importance due to the problems were investigated. The result of the study in the income structure but because of the was decided to use Analytical Hierarchy fact that it was one-to-one adaptation to Method. Thus, the main criteria that affect youth trends (Cinema tickets, innovative DSS the operator selection of persons and which services), the most important criterion was will be at the bottom of the hierarchical Extra Benefits. structure of AHP are determined. With the alternative operators in Turkey, namely Turkcell, Turkish Telecom and Vodafone, the

124

3. Prioritization of Public Services for functions are determined. At the same time, Digitalization the utilities of the digitalization that will be the criteria for the comparison are *Irem Ucal Sari, Istanbul Technical determined from the literature review. University, Turkey, Reduced cost, fast response, ease of [email protected] accessibility, reduced service times, increase Ozdemir Kucukali, Istanbul Technical in the available information and increased University, Turkey, quality are determined as the criteria. For the [email protected] determination of the criteria weights and comparison of the alternatives, Analytical Digitalization is one of the main focused Hierarchy Process is used. After strategies in organizations with the impact of determination of weights for these different Industry 4.0 revolution. The expectations and utilities gained by digitalization, departments the habits of the customers are rapidly and their services are evaluated according to changing as they gain more accessibility to the effects of their digitalization on the technology in their daily lives. This leads a big utilities. Health care services department is transformation on the customer service selected first for the digitalization as it has systems. By the utilization of IoT technologies the highest priority from the results. The in service systems, people are getting used digitalization of health care services involves to reach information easily. As well as the online consultation for the controls, online organizations, public sector should also initial diagnosis in emergencies, planning focus on digital transformation due to this scheduled visits and health consultations, change in the expectations of the citizens. ease of access to the information on the Public services could utilize from preventive health care and previous data digitalization to analyze consumption of share for agreed citizens. resources, decrease wastes and increase efficiency on the service time for different 4. A Decision Support Methodology activities. Additionally, digitalization will for Increasing the Efficiency of the provide a more detailed and larger data on Largest Border Crossing between citizens’ changing behaviors and their new Europe and Turkey expectations from the managers. On the other hand, citizens need to have Ilker Topcu, Istanbul Technical accessibility to the information they need University, Turkey, [email protected] using IoT tools and if it is possible, get the Füsun Ülengin, Sabancı University, service online. From that driving point, the Turkey, [email protected] main objective of this study is determined as Özgür Kabak, Istanbul Technical prioritization of the public services to decide University, Turkey, [email protected] which one should be digitalized earlier. To do *Şule Önsel Ekici, Doğuş University, this, first public services that could be digitalized are listed, and then the alternative Turkey, [email protected] tools, applications are discussed to Berna Unver, Sabancı University, determine alternative technologies, which Turkey, [email protected] will be compared in the study. Public Turkey is at the crossroads of the new Silk services that have high interaction with Road connecting China to Europe. This silk citizens and have problematic processes are road would be redeveloped as the Belt and the ones, which could utilize more from the Road Initiative, one of the most significant digital transformation. Therefore, health care infrastructure project of the new era. After services, waste disposal department, public the completion of this project, the current transportation, information services, social heavy traffic flow through Turkey will be care services, and citizen complaints extremely heavier due to international trade. resolution centers are selected for initial This is why, the efficiency of border crossings digitalization. It is believed that the between Turkey and Europe is of primary digitalization of these services will decrease importance. Turkish authorities, being aware the idle times and increase efficiency, by this of this new challenge, made important way, service processes will be improved. For efforts to improve the efficiency of border each of the department available platforms, crossing operations, especially at Kapıkule applications are examined and their main Border Crossing operations. Kapıkule Border 125

Crossing is the busiest land border crossing WED-2-P point in Turkey and also in Europe. This crossing is the second busiest in the world. It Plenary Session 3 has become inadequate to process the Wednesday 11:10-12:10 - Room: Mustafa increasing number of trucks transporting Kemal Amfisi goods between Europe and Turkey. To address this issue, in a previous study, a Chair: Murat Köksalan detailed process analysis was conducted, a simulation model was developed, and New Results for Solving Imperfect- several action plans representing Information Games improvement strategies were analyzed and *Tuomas Sandholm, Carnegie Mellon, ranked in terms of the daily average number United States, [email protected] of trucks waiting in the queue to enter and leave the border crossing. However, to Most real-world settings are imperfect- increase the efficiency of Kapıkule Border information games. They present challenges Crossing, the selection of the action plans beyond those in perfect-information games. should not be solely based on the average In 2017, our AI Libratus beat top humans in number of trucks but should also be the main benchmark, heads-up no-limit evaluated from a much broader perspective Texas hold'em. In this talk I will discuss some taking into account several objectives that of our more recent work on imperfect- are generally in conflict with each other. information games. Topics include a unified Therefore, in this study, we propose a multi- framework for abstracting games with criteria decision support methodology that bounds on solution quality [Kroer & evaluates these action plans by considering Sandholm, NeurIPS-18], a sound depth- additional attributes; such as investment limited search framework [Brown & cost, operations cost, sustainability, border Sandholm, NeurIPS-18], the fastest security, and the satisfaction of the equilibrium-finding algorithms [Brown & beneficiaries, establishing a preference Sandholm, AAAI-19], deep learning as an ranking of action plans to improve the alternative to abstraction [Brown & capacity of the Border Crossing. Our results Sandholm, Deep RL Workshop-18], a general have implications for policymakers not only framework for online convex optimization for in Turkey but also in the EU. In accordance sequential decision processes and with the multi-attribute decision-making extensive-form games [Farina et al., AAAI- framework, in this study, three main stages, 19], the first scalable algorithm for namely, problem structuring, decision trembling-hand equilibrium refinements modeling, and decision analysis, are [Farina et al., NeurIPS-18], and trembling- conducted. In the problem structuring stage, hand refinement of Stackelberg equilibria we identify the alternatives which are [Farina et al., IJCAI-18; Marchesi et al. AAAI- improvement action plans and the attributes 19]. which will be used for evaluation. In the decision modeling stage, based on the WED-3-P judgments of the stakeholders, the Doctoral Dissertation Awards importance of attributes is revealed. Additionally, we ask the stakeholders to Wednesday 13:10-14:25 - Room: Mustafa evaluate the performance of alternatives Kemal Amfisi with respect to the attributes identified in the Chair: Jose Rui Figueira first stage. Finally, in the decision analysis stage, we prioritize the alternatives, using a 1. Output-Sensitive Complexity of well-known and widely implemented multi- Multiobjective Combinatorial attribute decision making method: PROMETHEE using the Visual PROMETHEE Optimization Problems with an software. The prioritization of the Application to the Multiobjective alternatives is then presented to the Shortest Path Problem authorities *Fritz Bökler, Osnabrück University, Germany, [email protected]

126

In computational complexity of I investigate the practical effectiveness of multiobjective combinatorial optimization approximation algorithms for the MOSP (MOCO) problems, it is long known that even problem. It is a well-studied problem and the easiest problems in terms of single practically useful algorithms exist to some objective optimization do not admit a extend. Still, if the number of objectives is polynomial time algorithm in their larger than just a few, exact methods turn out multiobjective versions, because the number to be prohibitive. There are several known of Pareto-optimal outcome vectors may grow approximation approaches, but they all share exponentially in the size of the input. Thus, one downside: The running time can still be from the standpoint of classical complexity exponential in the number of objectives. theory, most MOCO problems are Moreover, while the worst-case is well indistinguishable. In my thesis, I establish controlled in these algorithms, the best-case the use of output-sensitive complexity theory- usually is not good. Especially in the case of --a concept well-known in theoretical the state-of-the-art-algorithm, the worst-case computer science---in multiobjective matches the best-case. In contrast, in exact optimization. The basic assumption is that label-setting algorithms, the best-case we want to compute the Pareto-optimal running-time matches the running time of a outcome vectors and thus accept the fact simple Dijkstra-Algorithm. So, in my thesis, I that the number of outcomes can be design a new approximation algorithm based exponential in the input size. But we still want on labeling algorithms to use the decades of an efficient algorithm to be fast if the knowledge we have on labeling algorithms. outcome is small and not too slow if the The goal is to construct it in a way so it is outcome is large. Hence, we look at the usable in practice and still gives the same running time as a function of the input and quality guarantees. While it turns out that its the output size. An algorithm is thus called worst-case running-time guarantee is not as efficient, if we can bound its running time by good as the state-of-the-art, I conduct a polynomial in the input and the output size. extensive computational experiments that An efficient algorithm in this regard is also show the practicability of my implementation called output-sensitive. To justify this new of the new approximation algorithm in complexity theory, I give easy and hard contrast to the existing approximations. It examples of multiobjective combinatorial turns out, that the approximation algorithms optimization problems: I prove that the from the literature can only solve very small multiobjective global minimum cut problem instances, while my new approximation is an easy problem in this regard, as well as- algorithm is much more usable in practice --under mild assumptions---multiobjective and is capable of solving real-world instances linear programming (MOLP) and finding of meaningful size. extreme nondominated points of MOCO problems. I investigate the latter approach in 2. Preference Disaggregation: an experimental study and show that the use Towards an Integrated Framework of a theoretical analysis framework can lead *Mohammad Ghaderi, Pompeu Fabra us to selecting and improving algorithms that are also very usable in practice. This also University (UPF), Spain, solves a central open question in [email protected] multiobjective combinatorial optimization, Preference disaggregation aims at capturing since a (practically) efficient computation of preference models by decomposing indirect the extreme nondominated points was only preference information. The preference known for the case of two and three model reflects on the conflicting points of objectives. To further justify output-sensitive view that collectively form a basis for the complexity, I also provide non-trivial hard judgments. This topic is addressed by several examples: I develop a methodology adopted disciplines, including behavioral science from the theory of output-sensitive (decision analysis), artificial intelligence complexity and apply it to multiobjective (preference learning), economics and optimization. I prove that the general marketing (choice modeling). These streams, multiobjective single pair shortest path although originated from different (MOSP) problem is hard in this regard and philosophies, are converging to a cannot be solved in an output-sensitive way comprehensive understanding of human unless P = NP. In a second part of my thesis, preferences as the main ingredient of 127 decisions. This dissertation sheds light on journal). The dissertation, as a compendium this phenomenon by introducing an of publications, takes a constructive view of integrated analytical framework that allows preferences and contributes to MCDA capturing preferences of a complex form literature by addressing three fundamental based on preference information of the components in three main aspects of simplest form. The research contributes to preference disaggregation paradigm. First, preference disaggregation paradigm in great potentials of MCDA frameworks for Multiple Criteria Decision Aiding (MCDA) by formulating and exploring complex business adding several methodological flexibilities problems, especially those without sufficient that are built on Linear Programming (LP) prior knowledge of the context, are techniques. In the first part of my thesis, I demonstrated. Second, an efficient introduce a motivating example based on the analytical framework for elicitation of case of nonmonotonic preferences in the preferences and construction of value brand color context, addressing how brand functions in presence of non-monotonicities color can influence customers perception of is introduced. Third, importance of decision a brand. To this aim, an iterative approach problem settings in preference based on LP technique is developed and disaggregation to derive robust applied to a real data set of major beauty & recommendations and construct expressive care brands (published in the Pattern models is studied. Recognition Letters journal). The next part of the thesis introduces a comprehensive 3. Portfolio Selection in Evolutionary analytical framework for capturing Algorithms preferences of complex form by processing indirect preference information in form of *Andreia P. Guerreiro, University of holistic comparisons. The methodological Coimbra, Portugal, [email protected] framework is designed also to be Evolutionary Algorithms (EAs) are algorithms computationally efficient so that can be inspired in the process of natural selection, employed in decision problems with large and are among the most popular methods in number of alternatives or numerous multiobjective optimization. As objectives are evaluation criteria, or even discretized typically conflicting with one another, instead continuous decision problems. The analytical of a single optimal solution, there is usually a framework is developed based on LP and set of optimal solutions which, together, form enables capturing complex preferences by a trade-off surface. The choice of the best exploiting preference structure, while solution depends on the preferences of a maintaining the balance between complexity Decision Maker (DM), which arise from and expressiveness of the constructed subjective information not contained in the preference model. The effectiveness of the multiobjective optimization problem methodology is demonstrated through formulation itself. In the absence of such comparison with the benchmark models and preference information, an optimization a large simulation analysis (published in the algorithm should be able to cover the optimal European Journal of Operational Research). tradeoff surface as well as possible, in order The third part of the thesis follows an to increase the chance that at least one extensive experimental analysis based on satisfactory solution is presented to the simulation to study the impact of decision (unknown) DM. One of the main steps of an problem settings and parametrisation EA is selection, which is aimed at focusing on procedures on the outcome quality of the the best solutions, but must also maintain a preference disaggregation analysis. The sufficient level of diversity in the population. study examines existing parametrisation A very successful selection approach procedures in various problem settings, and consists in optimizing a measure of the discovers conditions under which the quality of the population as a whole, existing procedures fail to provide robust and accounting simultaneously for individual expressive results. Based on these findings, quality and population diversity. Such EAs a new supervised parametrisation strategy is are called indicator-based EAs, as they rely introduced, which demonstrates a ”good” on set-quality indicators to perform selection. performance across heterogeneous sets of Indicator-based EAs are currently state-of- simulated settings (published in the the-art algorithms in Evolutionary Computers and Operations Research Multiobjective Optimization (EMO). More 128 recently, the notion of selection in EAs has been linked to the Portfolio Selection Problem (PSP), which is well known in Finance. In this analogy, individuals are seen as assets whose returns are random variables that are characterized by their expected values and covariance matrix. Balancing between good and diverse solutions in a population becomes analogous to balancing expected return and risk, respectively, in financial portfolios. In particular, it has been empirically shown that such a balance in EMO selection can be achieved using the risk-adjusted performance index known as Sharpe ratio, without modification, as a new quality indicator in the context of a particular formulation of random individual return related to the concept of dominated hypervolume. The focus of this work is the subset selection problem at the core of selection in EMO algorithms, but from the more general point of view of Portfolio Selection. Two lines of work are followed. The first one consists in improving those state-of- the-art algorithms that are based on the hypervolume indicator, which is a well known and theoretically supported, but computationally expensive, quality indicator. The second line of work goes beyond subset selection: a new type of indicators based on the Sharpe ratio is studied, both theoretically and experimentally, shedding new light on selection, fitness assignment and preference integration in EMO algorithms.

129

and unsupervised methods. In supervised methods, the decision maker inputs THURSDAY information to the optimization process. This intervention by the decision maker can be problematic in several scenarios. For THU-1-A instance, if the objectives are exogenously Contributed Session: Multiple Criteria imposed by regulatory parties as in the case Ranking or Sorting of safety related issues, the decision maker can be liable for the weights defined when Thursday 9:00-10:40 - Room: Galata Tower the weighting method is used. Therefore, in such scenarios, reduction to single-objective Chair: Jean Rosenfeld should be unsupervised to avoid bias 1. Comparison of Unsupervised towards the decision maker’s preferences. Methods to Reduce Multi-Objective There are several unsupervised methods for reducing a multi-objective optimization Models to Single-Objective Models problem to a single-objective one. In this *Dincer Konur, Texas State University, study, we discuss three unsupervised United States, [email protected] methods and compare them. Particularly, equal-weighting, min-max-deviation, and Samuel Vanfossan, Missouri University distance-to-ideal-point are the three of Science and Technology, United unsupervised methods considered. Under States, [email protected] equal-weighting, the objective functions are Two main approaches to solve multi- equally weighted; under the min-max objective optimization problems are (i) deviation method, the maximum deviation generating Pareto efficient solution from the individual objectives’ optimal values alternatives and (ii) reducing the multi- is minimized; and under the distance-to- objective optimization problem to a single- ideal-point method, the single objective is to objective one. The former approach returns minimize the distance to the ideal point (a alternative solutions, among which the point which has the optimal objective decision maker needs to select one to function values of the individual objectives). implement. Therefore, even if this approach We first compare these three methods using is used, the later approach would be utilized a set of randomly generated alternatives with in the final selection. On the other hand, in different number of objectives (this the later approach, a multi-objective comparison is motivated from hazardous optimization problem is reformulated as a materials routing on rail with many safety single-objective optimization problem, and criteria that are suggested to be considered then the corresponding single-objective by federal authorities). After that, our focus is model is solved to generate a single solution. on multi-objective binary knapsack problem. Especially for problems with many (more We analyze a set of problem classes (with than three) objectives, reduction to single different sizes and numbers of objectives) objective is needed even if Pareto efficient and compare these three different methods. solutions are generated because the number Our comparison focuses on robustness of of Pareto efficient solutions significantly each method as measured by how well each increases with the number of objectives method performs based on the other considered. There are different methods for methods’ single-objective function definition. reducing a multi-objective optimization 2. Financial Rating with Ordinal problem to a single-objective one. For instance, weighting (ranking) different Classification based on the objectives and formulating a single objective Hierarchical SMAA Choquet Integral function as the weighted sum of the Approach individual objective functions is one such method. However, in most weighting *Sally Giuseppe Arcidiacono, University methods, the decision maker's preferences of Catania, Italy, [email protected] are sought, i.e., such methods are Salvatore Corrente, University of supervised by the decision makers. The Catania, Italy, single-objective reduction methods can be [email protected] classified into two categories: supervised 130

Salvatore Greco, University of Catania, degree of sustainability achievement while Italy, [email protected] evaluations are uncertain and temporal? Are existing sorting multi-criteria methods could We consider financial rating proposing a be of used ? This paper aims to answer robust ordinal regression and a stochastic these questions. It proposes a generalization ordinal regression model taking into account of SMAA-Tri to temporal context (multi-period interaction in a hierarchy of criteria evaluation of alternatives). The proposed corresponding to financial and economic method will allow the sorting of each decision indicators underlying the rating. Formally we depending on its temporal and stochastic apply Robust Ordinal Regression (ROR) and impacts. The results will assign each Stochastic Multiattribute Acceptability decision to a category according to their Analysis (SMAA) to a sorting model based on sustainability level under climate change the Choquet integral taking into account a uncertainties. Temporal evaluations are hierarchy of criteria. The advantages of our represented by immediate and future approach are shown by means of an impacts of decisions on a finite horizon. application to financial rating of European Stochastic uncertainty is modelled with Countries. Economic and financial data used Monte-Carlo simulations as in SMAA Tri in the analysis are provided by Standard (Tervonen et al., 2009). The proposed &Poor's Global Inc. method consists of two phases i) a multi- 3. A Temporal Sorting Multi-Criteria criteria aggregation and ii) a multi-period acceptability index aggregation. Multi-criteria Approach under Stochastic aggregation consists of applying SMAA Tri at Uncertainty each period. As a result, acceptability index Youness Mouhib, Université du Québec is computed for each period and for each category. The second phase will consist in an à Rimouski, Canada, aggregation of the acceptability indices [email protected] obtained over the multi-period horizon. We *Anissa Frini, Université du Québec à will make an analogy considering each Rimouski, Canada, category as an alternative, each period as a [email protected] criterion, and each relative importance of periods as the criteria weights and apply In recent years, Quebec government has TOPSIS. TOPSIS will provide us with a score highlighted the importance of making for each category representing the closeness sustainable decisions that take into account to the ideal solution. The ideal solution is a the sustainable development principles. category for which the acceptability index is Furthermore, climate change and the equal to 1 at each period. This research work associated uncertainties has been added as demonstrates how the paradigm behind a major concern in the last few years. outranking sorting methods and specifically Consequently, decisions need to be both SMAA-Tri, can be generalized to the context sustainable and robust under climate of temporal evaluations of decisions. The change uncertainties. In this context, the proposed approach is then applied in need for new multi-criteria decision-making sustainable forest management context. The methods, which are able to take into account objective of the illustration is to sort each immediate and future impacts in a long-term forest management option according to its perspective and which model the uncertainty sustainability achievement and climate related to climate change vulnerabilities, has change uncertainties. First, we will need to emerged. Such methods will support balance several standpoints, such as decision-making in all government biodiversity conservation, soil and water departments and agencies and in different conservation, maintenance of forest domains such as water management, ecosystem productivity, maintenance of electricity generation, spatial planning, multiple socio-economic benefits and healthcare, transportation, manufacturing, consideration of the population’s values and energy, forest management, etc. Several needs. Second, the option is evaluated in the questions arise in this context and deserve to short, medium and long term (150 years be answered. How to evaluate the which corresponds to forest regeneration sustainability of decisions? Can we sort period). Third, evaluations and criteria decisions into categories according to their weights will be considered stochastic and

131

Monte-Carlo simulation will be done in order to assess the quality of a partition. This to provide a robust sorting of the decision. indicator is based on the asymmetric preference relation that we use in the References: Tervonen T., Figueira J.R., multicriteria k-means extension. Lahdelma R., Almeida Dias J., Salminen P., Experiments on an artificial dataset, (2009) “A stochastic method for robustness consisting of clusters designed to present a analysis in sorting problems”, European partial order based on two criteria, illustrate Journal of Operational Research, vol 192, pp. the results provided by the proposed method 236-242. and compare them with the standard k- 4. An Alternative to L1 K-Means means ones. We then consider the World Algorithm to Obtain Partially Ordered Happiness Report 2018 dataset and show that the proposed method gives better Clustering results for the quality indicator based on the *Jean Rosenfeld, Université libre de preference relation, while the standard k- Bruxelles, Belgium, [email protected] means method is better for the Dunn index. Yves De Smet, Université libre de The main objective of this work is to present a common framework to stimulate Bruxelles, Belgium, discussions between experts in data mining [email protected] and multicriteria decision aid. Christine Decaestecker, University of Brussels (ULB), Belgium, THU-1-B [email protected] Contributed Session: AHP/ANP Clustering problems are well known in data Applications in Aviation mining. Methods, such as k-means or Thursday 9:00-10:40 - Room: Maiden's hierarchical algorithms, are widely used in Tower practice. These approaches generally use a distance metric to form groups that are as Chair: Raimo Hämäläinen homogeneous as possible and as heterogeneous as possible between 1. Multicriteria Decision Making themselves. In our opinion, traditional Based on Sustainability in The clustering techniques cannot be directly Selection of UAV Applications for applied in a multicriteria context. Indeed, Smart Cities since criteria have to be optimized, the comparison between two alternatives usually *Mustafa Hamurcu, Kirikkale leads to asymmetric relations (in opposition University, Turkey, to the symmetric relation provided by a [email protected] distance metric). This distinctive feature Tamer Eren, Kırıkkale University, Turkey, opens new perspectives for clustering such [email protected] as the establishment of preferential relations between clusters. Totally ordered Smart city is defined as an advanced modern multicriteria clustering consists in detecting city that utilizes information communication a partition where a complete and transitive technology and other technologies.The smart relation exists between each pair of clusters. city applications can improve both quality of This means that the first cluster is life, operational efficacy of urban services considered better than the second, which is and ensuring the resource availability for better than the third one, etc. Several present and future generations in terms of multicriteria clustering models have been social, economic, and environmental developed over the past ten years, including aspects. The smart city emerged as a partially ordered multicriteria clustering solution to address the challenges arise due where some cluster pairs cannot be ordered. to exponential growth of urbanization and In this contribution, we investigate a minimal population and its popularity is increasing extension of the well-known k-means day by day.Smart cities can elevating the algorithm to address multicriteria problems. level of comfort of residents with activity This will allow us to build partial relations such as optimize resources utilization and between the clusters. A quality indicator is enhance operations in health, also adapted from the standard Dunn index transportation, energy, and environment. One of smart city technologies is the 132 unmanned aerial vehicles (UAVs), which can evaluation model based on the FAHP and the provide many applications for smart cities TOPSIS to help the actors in the military and create a positive impact on the society. sector for the selection of optimal UAV. In There are various UAV applications such as order to illustrate the use of the model for the traffic monitoring and management, security UAV selection problem, an application is and crowd monitorings, UAV-based made with UAVs used by the air forces of infrastructure inspections, health emergency some countries. In this methodology by services and UAV taxi. This paper include utilizing improved AHP by fuzzy set theory, decision making process for selection of first we try to calculate weight of each smart applications in smart cities. This criterion. Then by implementing TOPSIS decision process is applicated for Istanbul algorithm, assessment of UAVs has been metropolitan municipality and the best done to obtain final ranking. To determine selection/ranking is made and ranking of the most appropriate UAV to help the actors preference are made for the most in defense authorities, first, evaluation practicable UAV applications. In this decision criteria affecting operator preferences are making process, we use multicriteria identified based on a detailed literature decision making methods namely fuzzy review and expert opinions. The authors analytic network process (ANP) based on come up with a set of main criteria that sustainability criteria using fuzzy numbers. includes maximum payload, operational radius of action, cruise speed, maximum 2. Unmanned Aerial Vehicle speed, service ceiling, takeoff weight, Selection Using Fuzzy AHP and powerplant, wingspan, endurance, and TOPSIS Methods length. After constructing the hierarchy, the authors ask the operators of UAV pairwise *Erhan Berk, National Defense comparison questions to assess the relative University, Turkey, [email protected] priorities of evaluation criteria. UAV Emine Egdemir, National Defense operators are taken as the target group of University, Turkey, the survey. Twenty UAV operators participate [email protected] the survey. As a further study, survey results can be generalized by expanding the The military usage of unmanned aerial participant group and interacting with other vehicles (UAVs) has risen as a direct need of staff involved in the operation of the UAV the armed forces. The UAVs provide ease of system. operation to the armed forces, particularly in the border regions, providing intelligence, 3. Application of Analytic Network surveillance and reconnaissance (ISR) as Process to the Interview Examination well as a strike capability to understand and in an Air Traffic Control Department defeat the threat. The UAVs offer a significant advantage by avoiding operator damage *Mustafa Özdemir, Eskisehir Technical while providing the same and in some cases University, Turkey, superior capabilities as a manned aircraft. [email protected] For example, some UAVs can stay aloft for Mujgan Sagir, ESOGU - Eskişehir longer durations than expected on crewed Osmangazi University, Turkey, aircraft. UAVs play a vital role in present day operations due to their asset capabilities, low [email protected] operating costs and their ability to reduce the In the aviation sector, air traffic controllers risk of life of pilots. In this paper, by using (ATCOs) and pilots are the staffs who have fuzzy analytical hierarchy process (FAHP) and the high responsibilities in flight operations, the technique for order performance by and they undertake critical tasks where even similarity to ideal solution (TOPSIS) a small error may cause serious technique we propose a new method for UAV consequences. ATCOs are responsible for selection problem. The problem of UAV ensuring the safe, regular and rapid flow of selection is a strategic and significant issue, aircraft traffic in airspace and at airports. especially in the military sector. Moreover, They consistently carry out their duties in co- selecting the optimal UAV among many ordination and co-operation with pilots, alternatives is a multi criteria decision technical personnel, management and other making problem. This paper presents an supervisors. Pilots provide take off, landing,

133 descent and other maneuvers of aircraft in In the early decision making literature accordance with the instructions received intuitive and fast decision making was from ATCOs. Someone who is not appropriate studied under the theme naturalistic for this profession is more likely to make decision making but the related questions mistakes when they are on duty. At this point, seem to have been forgotten today. the capabilities of the ATCOs become crucial. Situational awareness (SA) and team The selection of ATCOs and the training situational awareness also relate to fast quality given after selection should be decisions and were studied early in focused so that skilled and qualified people connection with pilot decision making . These can do this job. Failure to select a suitable topics have increasing practical relevance candidate is a major obstacle to training a today. In many areas decision making is qualified ATCO, no matter how good the done in teams and also in a rapidly changing quality of the training is. In the study, firstly, environments with multiple criteria. It is the process of admission of candidates to an characteristic that there is not time for air traffic control department was examined. extensive analyses and model development In the current process, students should take at the time decisions have to be made. Such a series of examinations including areas include, e.g., emergency situations, mathematics, aptitude, interview and fire rescue, acute medical operations as well simulation. These examinations, except the as rapid prototyping in project development. interview, are more convenient for We demonstrate the challenges in rapid measurable and objective evaluation. team decision making by describing a However, in the interview stage, candidates military example of decision making in air are subjected to a more qualitative combat by a team of fighter pilots. In these evaluation by an interview committee. A decision situations behavioral issues certain number of the candidates listed become most important. The way the team according to the interview scores are entitled interacts and shares the situational to proceed to the final stage. To improve the awareness and the mental models of the evaluation process, we consider this phase overall team becomes essential. Originally as a multicriteria decision making problem the literature on SA mainly considered with tangible and intangible criteria and military decision making and the focus was propose an Analytic Network Process (ANP) on the related cognitive factors. However, in model. ANP is able to determine criteria teams emotions and social dynamics as well weights considering dependency and as the creation of trust can also be feedbacks among them. Multi-criteria important. It is well known that emotions decision-making methods for personnel play a crucial role in decision making. Thus, selection in the air traffic control is a new it is natural to suggest that socio-emotional area, and only few studies are available in factors should also be taken into account the literature. Five main criteria are when evaluating the situational conditions in determined, through a survey, as team decision making. We propose that the communication skills, stress control, concept of Systems Intelligence (SI) can be attention level, job interest and self-reliance. a way to bring these missing human The evaluation of the candidates for each elements into the generalized definition of criterion is made using rating method due to SA. SI emphasizes the importance of large number of alternatives. Finally, the systems thinking and the attunement among ranking obtained by ANP and rating method the team members. Our suggestion is that is compared with the current system. including SI as a topic in the training of the members in the decision making team will 4. Team Situational Awareness and help in generating trust and shared SA which Systems Intelligence in Fast Group again are a prerequisites for good decisions. Decision Making Rapid decision making relates to intuitive and heuristic processes which reflect *Raimo Hämäläinen, Aalto University, System 1 thinking in the dual processing Finland, [email protected] theories of human cognitive behavior . The Kai Virtanen, Aalto University, Finland, evaluation processes carried out e.g. in [email protected] traditional multi-criteria analysis rely on Heikki Mansikka, Aalto University, System 2 thinking. So one could say that the Finland, [email protected] challenge in supporting rapid decision 134 making is to develop procedures which comparisons between criteria or would enable the decision maker to use alternatives. The mainstream research System 1 processes to intuitively or focus on the ambiguity in this process. Fuzzy heuristically generate solutions in different logic is used to overcome this problem. On contexts with a multi-criteria perspective. A the other hand, there is another challenge to natural first idea to reach this goal is to train cope with to make accurate decisions: intuition by practicing multi-criteria Cognitive biases. Thanks to pioneer study of evaluations beforehand in different Nobel winner researchers Tversky and situations related to the context at hand. One Kahneman (1974), bias in human way of approaching this is to develop judgement and decision making is prototype practices which is typical, e.g. ,in discussed. It was shown that people rely on fire emergencies . In fighter plane pilot a limited number of heuristic principles training the practices refer to tactics, which reduce the complex tasks to simpler techniques and procedures . In these judgmental operations. But these heuristics practices the multi-criteria dimensions of the lead to severe and systematic errors by the decision remain implicitly embedded in the way they are quite useful. The first three good practices. There is no time for explicit heuristics described in the first article are analysis. In multi-criteria decision making representativeness, availability of instances one can practice decision in different and adjustment from an anchor. Gains and situations with generic pre-structured value losses are evaluated differently due to the tree models . This is also applicable for shape of the value function, which is concave groups .The group can jointly develop the for gains and convex for losses according to value trees for different situations and the Prospect Theory (Kahneman & Tversky, practice the approach beforehand. This has 1979). Tversky and Kahneman (1981) also already been considered in nuclear demonstrated choices can be influenced by emergency preparedness exercises . The the framing (i.e. wording) of decision current literature on multi-criteria group alternatives. Two decision-making processes decision support has, as far as the authors (System 1 and System 2) in the brain is are aware of, not considered rapid decision explained in “Thinking Fast and Slow” book making nor the concept of shared SA. When written by Daniel Kahneman. System 1 the emotional elements are included in SA, thought is rapid, instinctual and emotional, the new perspective can help to understand often on impulse and is part of the fight or the systemic nature of participatory group flight response. Contrary to System 1, decision making in different domains ranging System 2 thought is slower, more deliberate from life threatening situations to value and logical. According to the research on laden contexts such as environmental cognitive biases, it is taught to research of decisions. the effect of cognitive biases on MCDM. It was found in the literature that anchoring THU-1-C effect in interactive MCDM is studied by Contributed Session: Contemporary Buchanan (1997). However, the second Approaches in MCDM study found in the literature is published after twenty years in 2017 by Borrero and Thursday 9:00-10:40 - Room: Dolmabahçe Hanao. The sentence of “The relationship Palace between cognitive biases and these decision-making tools has not been amply Chair: Vasileios Mantogiannis investigated.” shows the need of research on 1. Cognitive Biases in MCDM this issue. Two motivational biases, Methods: An Embedded Filter respectively known as confirmation and desirability of choice are studied in that Proposal research. The efforts to diminish the effects *Nurcan Deniz, Eskişehir Osmangazi of biases is called “debiasing” (Cheng and Univesity, Turkey, Wu, 2010). It is aimed to propose a filter to [email protected] counteract the accurate decisions about cognitive biases as a debiasing strategy in The backbone of the Multi Criteria Decision this study. The first filter is restructuring the Making (MCDM) techniques is the specialist questions not to cause biases. Einstein’s “If I assessment. The specialist make pairwise were given an hour in which to do a problem

135 upon which my life depended, I would spend attitudes towards risk and psychological 40 min studying it, 15 min reviewing it and 5 biases. Behavioral experiments have proven min solving it” quote shows that the difficult that such approaches represent the real part of decision making is often not about the decision making process only poorly and right answer, but rather asking the right reduces the accuracy and reliability of the question (Ryan, 2018). The other filter is decisions. A review of the literature shows manipulating the weights according to the that different theories have been developed degree of expected cognitive biases. For to address a particular set of these issues. example it is proved that women decisions The most common ones are fuzzy set theory affected more than men decisions from and prospect theory. Although fuzzy set cognitive bias. There will be produced a theory has been successfully integrated with debiasing factor to multiply the weights. Also various MCDM techniques, the same is not from the point of view that deciding faster will necessarily true for prospect theory. In increase the bias risk, the response time will recent years, there has been a growing be measured and will be used to manipulate interest in these theories, especially in the weights (Guo, 2017). Expertise level is prospect theory, which is a well-established also will be quantified to manipulate the descriptive theory concerned with how weights. The other thing need to be handled humans actually make decisions and how is to look at the objective of the criteria they behave under risk. Reference (minimizing/maximizing) to detect loss dependence, loss aversion and nonlinear aversion affect. This issue is very important weighting of probabilities are key features because it is proved that clinician’s decisions introduced by this theory to explain the are affected from cognitive biases (loss judgmental principles that govern the aversion bias, numeracy bias toward evaluation of alternatives and the weighting frequencies omission bias, confirmation bias of uncertain outcomes. The theory departs and availability bias) (Braverman and Barby, from the classical assumption that decision 2012). Pandey and Jessica (2018) also makers are completely rational and provides measured behavioral biases affecting real a better understanding of the factors estate investment decisions in India. Shu influencing the decision process. Thus, (2018) studied in her PhD thesis about hiring integrating prospect theory into MCDM would a personnel bias effect.These application yield more accurate and reliable decision areas show the importance of this issue. making models. However, this requires addressing some issues related to data 2. A Behavioral Approach to MCDM collection and evaluation as well as group Problems decision making. A number of studies have been suggested to handle these issues, but *Ozgur Yanmaz, Istanbul Technical none of them are sufficient. In this study, an University, Turkey, [email protected] approach is introduced that extends the Umut Asan, Istanbul Technical application of prospect theory to MCDM University, Turkey, [email protected] problems under uncertainty. A decision Multi-criteria decision making (MCDM) is maker’s risk attitude (subject to a certain primarily concerned with modeling and reference point) is determined and analysis of judgments about alternatives integrated into a multi-criteria group decision with respect to multiple criteria. MCDM often framework to select the best alternative. The involves dealing with uncertain information proposed approach aims to improve the as well as highly subjective judgments of a accuracy of decisions by studying additional group of individuals. Judgments are affected dimensions of the decision-making process. not only by factors related to the problem but It introduces methods for data collection and also by the context within which the aggregation of individual judgments. In order judgment is made and by individuals’ past to demonstrate the effectiveness and experience, beliefs, preferences and applicability of the proposed approach, an expectations. In spite of this fact, MCDM MCDM application involving non-dominating techniques unrealistically assume that alternatives and conflicting criteria is decision makers are completely rational; in presented. Comparison with the results of other words, they ignore the fact that expected utility theory demonstrates that the humans’ cognitive abilities are limited and proposed approach provides a better that their decisions are influenced by their modeling of MCDM under uncertainty. 136

3. Combining Health Technology proposed decision-making model. Results of Assessment (HTA) & Multi-Criteria the proposed evaluation framework will be utilized by healthcare providers in the field to Decision Making (MCDM) Analysis to improve the quality of healthcare service and Evaluate Kidney Stone Treatment may give insights to all interested Alternatives stakeholders in the healthcare delivery system. *Beyza Özlem Yılmaz, Istanbul Medipol University, Turkey, 4. A Multi-Criteria, Multi-Stakeholder [email protected] Decision Aid Tool for the Assessment Eren Erol, Istanbul Medipol University, of Real-Estate Investment Turkey, [email protected] Alternatives under Financial Melis Almula Karadayı, Istanbul Medipol Performance Uncertainty University, Turkey, [email protected] *Vasileios Mantogiannis, Mott Hakan Tozan, Istanbul Medipol MacDonald, United Kingdom, University, Turkey, [email protected] [email protected] Fotios Katsigiannis, Lancaster University Management School, United Increasing prevalence of kidney stone Kingdom, disease during the last decade has pioneered the development of health [email protected] technologies that provide new treatment Investment decisions in private real-estate approaches to healthcare providers and demand the consideration of several patients. In this regard, applying a qualitative and quantitative criteria, as well convenient, highly accurate and low-cost as the different or even conflicting interests treatment approach became an important of the participating stakeholders. concern. In today’s healthcare, Meanwhile, certain indicators are subject to Extracorporeal Shock Wave Lithotripsy severe uncertainties, which will eventually (ESWL) and Laser Lithotripsy (LL) are alter the expected outcome of the accepted as effective and minimally invasive investment decision. Even though multi- procedures for fragmentation of kidney criteria decision making (MCDM) techniques stones. Pros and cons of these treatment have been extensively used in real-estate methods are still being questioned by investment appraisals, there is limited healthcare experts, which lead to the evidence from the private rented sector, necessity of a comparative analysis. At this which constitutes a large part of the existing point, MCDM is a powerful and convenient real estate assets. Existing approaches are tool for such kind of a comparative study. not designed to capture the inherent This study aims to compare ESWL and LL as variability of the decision environment while kidney stone treatment alternatives by at the same time, they do not always achieve hierarchical fuzzy TOPSIS which can handle a consensus among the participating actors. crisp and fuzzy data expressed in linguistic In this work, through a rigorous literature terms. A hierarchical evaluation structure review, we are able to identify an exhaustive consisting of 5 main criteria, namely, list of assessment criteria, which is further “Technical & Technological”, validated through an iterative Delphi-based “Organizational”, “Economic”, “Ethical, Legal consensus-making process. The selected & Social” and “Clinical” and 24 sub-criteria is criteria are then used to construct an constructed to determine the most suitable Analytical Hierarchy Process (AHP) model treatment alternative in an integrated evaluating four real-world, real-estate manner. A group of experts, consisting of investment alternatives from the UK private medical doctors (nephrologists and rented market. The volatility of the financial urologists) and researchers evaluate the performance indicators is grasped through importance of each criteria and sub-criteria several Monte Carlo simulation runs. We test and also the ratings of alternatives with the described solution approach with respect to sub-criteria. A case study will be preference data obtained by seven senior carried out at one of the biggest private real-estate decision-makers. Our hospital complex in Istanbul to apply the computational results suggest that financial 137 performance is the main group of selection between scenarios. In this work, we start criteria. However, the sensitivity of the from the perspective of a DM facing a outcome indicates that location and property scenario-based multiobjective optimization characteristics may greatly affect real-estate problem and consider some fundamental investment decisions. questions that a DM may expect to be answered by a suitable visualization. THU-1-E Examples of such questions are “which Contributed Session: Bringing objective values can be reached in a Multiobjective Optimization to predefined percentage of scenarios?” and “how to compare trade-offs between Applications objectives under conditions of different Thursday 9:00-10:40 - Room: Hagia Sophia scenarios?” We also classify problems into several cases based on the number of Chair: Nader Azad objectives, scenarios and solutions (e.g. a bi- 1. Visualizations to Support objective problem with few scenarios and few solutions to compare, or a problem with Scenario-based Multiobjective many solutions and/or many scenarios). Optimization Then, we propose different visualizations *Babooshka Shavazipour, University of that are well suited for each case and help Jyvaskyla, Finland, the DM answer the questions identified above. For example, we extend the empirical [email protected] attainment function for bi-objective problems Manuel López-Ibáñez, University of and heatmap visualization for decision Manchester, United Kingdom, making in higher dimensions. Moreover, the manuel.lopez- pros and cons of utilizing the proposed [email protected] visualizations are discussed and illustrated Kaisa Miettinen, University of Jyvaskyla, with several examples. To the best of our Finland, [email protected] knowledge, this is the first study in which visualization methods are particularly Uncertainty is present in almost all real proposed for scenario-based multiobjective applications, which further increases the optimization, which is quickly becoming an complexity of multiobjective optimization attractive area of research both in theory and problems. In the absence of reliable real-world applications. This research is a probability distributions that model this part of developing an open source software uncertainty, scenarios represent plausible framework DESDEO (desdeo.it.jyu.fi) for future states in which the same solution or interactive multiobjective optimization decision may present different objective methods. values. In scenario-based multiobjective optimization problems, the performance of a 2. A New Approach for Solving decision should be evaluated according to all Multiple Objective University Course objectives under the conditions of each Timetabling Problem plausible scenario. Thus, scenarios introduce an additional dimension to the *Mustafa Bayar, Ankara Haci Bayram solution process and complicate the task of Veli University, Turkey, the decision maker (DM). Visualizations can [email protected] help a DM to understand and compare not Murat Atan, Ankara Haci Bayram Veli only trade-offs between different objective University, Turkey, [email protected] functions but also trade-offs under different Irmak Uzun Bayar, Hacettepe University, scenarios. In traditional multiobjective optimization,visualizations are typically used Turkey, [email protected] for representing Pareto optimal solutions in Abstract Many scenarios, especially those the objective space, helping DMs to observe, that interest multiple decision-makers (DMs), learn and evaluate trade-offs between include multiple and conflicting objectives. In objectives and eventually choose the most such cases, the DMs generally provide a set preferred solution. In the presence of of preference information to sort and scale multiple scenarios, those traditional the satisfactory levels for the multiple visualizations fail to capture the trade-offs objectives. A broad range of conventional

138 multiple objective decision making (MODM) time. First, we propose a (2, 2)- approaches take only cardinal preference approximation algorithm for the information (weights) into account by simultaneous minimization of makespan and employing them in the objective function as flow time. The idea of the algorithm is the a weighted sum of the multiple objectives. following: -maximize the number of jobs put The weighted sum may not sort/scale the in a batch by initially sorting jobs in non- satisfactory levels with respect to the weights decreasing order of sizes, - if a job does not as well as being a radical function, that is the fit in an existing batch, job is split and its first satisfactory levels may be unresponsive to part entirely fills the open batch, - then a new small changes in the weight vector. In this batch is opened and the second split part is study, a new fuzzy multiple objective linear placed in that batch, - afterwards, all split programming (FMOLP) formulation is jobs are removed from batches and, if presented that address the aforementioned necessary, new batches are generated for issues and, that is able to process ordinal those jobs. Then, we develop a procedure to and cardinal preference information generate an approximate Pareto set simultaneously. The presented formulation is following the initial solution given by the applied to a real-life university course approximation algorithm. The Pareto set timetabling problem to reveal its advantages approximation procedure generates over the conventional approaches. additional batches to the detriment of maksepan and rearranges the assignment of 3. Generating a Partial Approximate jobs to have a better flow time value. Each Pareto Set for Makespan and Flow time a new batch is generated, the procedure Time Minimization: Case of an looks up the job having the largest size and External Sterilization Service places that job in the new batch. Then, the space emptied with the displacement of that *Onur Ozturk, University of Ottawa, job is filled with other jobs which are initially Canada, [email protected] scheduled at a greater instant. This process continues until the used capacity of the A sterilization service is intended to sterilize additional batch is maximized. We ran any medical device used in a surgical several numerical tests to test the procedure. Once sterile, these devices can performance of the approximation algorithm be reused in other surgeries. They are and the Pareto set approximation procedure. therefore called "reusable medical devices We also benchmarked the approximation (or shortly RMDs)". In the case of an external algorithm to a genetic algorithm from the service, RMDs used in different surgeries are literature, before and after implementing the sent to sterilization altogether at the same Pareto set approximation procedure. time. The steps of the sterilization process Numerical results showed that the are the following: rinsing and washing, approximation algorithm dominates the verification, conditioning, and finally, genetic algorithm for small and medium size autoclaving. After all these steps, RMDs are instances containing up to 20 jobs. On the ready for reuse. Among these steps, the other hand, implementation of the Pareto set washing is usually a bottleneck for the whole approximation procedure improves the initial process. Washers are capable of treating solutions given by the approximation many RMD sets (RDMs used in different algorithm and the benchmarked method by surgeries) at the same time. Thus, we model around 10%. the washing step as a batch scheduling problem where RMDs are treated as jobs 4. A Bi-Objective Optimization Model having different sizes. We focus on the for Planning Rail Hazmat Shipments minimization of total time of daily washing in the Presence of Random operations and work-in process levels. The importance of the first objective is due to Disruptions operational hours of sterilization services Armin Jabbarzadeh, École de which can be typically open ten hours a day. technologie supérieure de Montréal, The second objective aims to minimize the waiting time of RMDs just before being Canada, [email protected] washed. In the scheduling literature, those *Nader Azad, University of Ontario objectives are called makespan and flow Institute of Technology, Canada, [email protected] 139

Manish Verma, McMaster University, ensure that the risk measure remains within Canada, [email protected] tolerable limits, is developed. The complexity of the resulting optimization program, and Hazardous materials (hazmat), though the inherent structure, motivated deploying harmful to humans and the environment, are an augmented ε-constraint technique to integral to industrial lifestyle and thus need generate a formulation involving single to be transported in significant volumes. objective. The proposed analytical approach Railroad is one of the safest modes for was used to study the transportation system transporting hazmat, in large part because of of a Class I railroad operator in North the implementation of a comprehensive America, and the resulting analyses led to safety plan and a host of industrywide the following conclusions. First, rail hazmat initiatives, however, the possibility of risk can be reduced by using proactively spectacular events resulting from multi- adding extra capacity to the safer links, railcar incidents do exist. The derailment and/or reactive strategies of re-routing and and explosion of several crude oil rail tank using third party logistics. Second, the new cars causing irreparable destruction and loss risk measure that combines expected risk of human life in Lac-Megantic (Quebec, and variability in risk could be a very valuable Canada), in July 2013, is an example of the tool for both the railroad industry and the possible catastrophe associated with rail regulators who deal with low probability – hazmat shipments. Disruptions, induced by high consequence hazmat events. Third, any nature such as hurricane Kartina in 2005 or given railroad network can be preprocessed man-made threats such as the 9/11 terrorist to ascertain the critical and near-critical attacks in the United States, could either service legs for adding extra capacity to be threaten the transportation networks or commensurate with the risk attitude of the render a single transportation mode such as decision maker. Fourth, cost savings is railways unavailable. Unfortunately, the higher from adding extra capacity vis-à-vis re- existing risk management techniques have routing railcars. Fifth, safer service legs are limited efficacy in the event of disruptions, chosen for capacity addition even in the and the resulting unavailability of rail-links. absence of disruptions, because that would Thus, there is a need to study disruption to help cope with disruptions and also reduce railroad operations, especially since such hazmat risk under normal operating events are not infrequent. It should be conditions. evident that the resulting risk management methods and optimization techniques THU-1-F should not only incorporate disruptions but also prescribe contingency plans for rail Contributed Session: Practical hazmat shipments. However, implementing Applications of MCDM such techniques so that the transport risk Thursday 9:00-10:40 - Room: Basilica and the transport cost are within reasonable Cistern bounds under both normal and disruption situations is challenging because of two Chair: Matthias Ehrgott reasons: first, both the risk and the cost 1. MRP-WES: Multiple Reference elements would tend to increase during disruptions since safe and economic routes Point based Weak and Strong might not be available; and second, it is Synthetic Indicators difficult to evaluate the effectiveness of *Samira El Gibari, University of Málaga, contingency plans because of the inherent trade-off between upfront cost to the railroad Spain, [email protected] operator versus the possible mitigation (of Francisco Ruiz, University of Málaga, risk and cost) resulting from the disruption. Spain, [email protected] We make the first effort to fill the gap by José M. Cabello, University of Málaga, developing a bi-objective two-stage Spain, [email protected] stochastic program that considers random Trinidad Gómez, University of Málaga, disruptions in the tactical planning of rail Spain, [email protected] hazmat shipments. A new measure of hazmat risk that combines expected risk with In this paper, a novel methodology based on the variability in risk, and whose goal is to the multicriteria reference point scheme (the

140

Multiple Reference Point Weak-Strong The problem of finding the beam angles Synthetic Indicator, MRP-WES) is proposed. configuration (BAO) that leads to the best The decision maker can establish any treatment plan in radiation therapy is a very number of reference levels for each challenging optimisation problem. Further, indicator, and the final outcome can be as we aim to deliver radiation to a tumour interpreted in terms of the position with according to some medical prescription and, respect to these levels. In order to illustrate at the same time, minimise the damage on the behaviour of the scheme proposed, we surrounding organs/tissues, the problem is apply it to the construction of the EU- inherently a multi-objective one. To solve this Regional Social Progress Index, taking into problem, several strategies have been account the three dimensions considered: proposed in the literature. Among them, “Basic Human Need” (BN), “Foundations of multi-objective local search algorithms Wellbeing” (FW) and “Opportunity” (O). (MOLS) have been shown to be a very simple Besides, two different aggregations are and effective strategy to solve the BAO proposed: the weak indicator (WSI), allowing problem. However, MOLS algorithms have for full compensation among the single several problems we need to address to indicators, and the strong indicator (SSI), not think of applying this kind of strategies in allowing for any compensation. If we wish to clinical practice. One main problem is their propose a ranking, partially compensatory slow convergence. As mentioned previously synthetic indicators (MSI) can be built for in the literature (Cabrera-Guerrero, 2018), different compensation degrees, ranging MOLS can take many hours (or even days) from 0 to 1. From the practical point of view, before to converge to a set of locally efficient the main advantage of MRP-WES is twofold. treatment plans. Further, most of these First, the MRP-WES indicators can be easily plans tend to concentrate on the same part interpreted as the global position of the of the objective space, leading to a lack of corresponding unit with respect to diversity among the locally efficient solutions hypothetical global reference levels, found by the algorithm. In this study, we resulting in a meaningful measure. Second, propose to use the epsilon-dominance apart from providing an overall measure, the concept to reduce the number of non- MRP-WES method provides warning signals dominated points the MOLS keeps in its which let the decision maker detect bad archive at each iteration, so the algorithm performances in certain indicators that may can converge faster. Also, we implement remain unnoticed otherwise. Finally, we some rules to improve the diversity of the show that the MRP-WES indicators satisfy a efficient treatment plans in the archive at series of properties which are regarded in the each iteration. We apply these ideas to a scientific literature as highly desirable for a Pareto Local Search (PLS) algorithm Synthetic indicators. previously proposed in the literature (Cabrera-Guerrero, 2018). We consider two 2. Accelerating Pareto Local Search different neighbourhood definitions. The Algorithms using Epsilon Dominance algorithm is tested on two clinical prostate for the Beam Angle Optimisation cases. While the tumour is located on the Problem prostate, organs at risk considered in this study are the bladder and the rectum. *Guillermo Cabrera-Guerrero, Pontificia Preliminary results show that using the Universidad Católica de Valparaíso, epsilon-dominance rule allows us to reduce Chile, [email protected] the number of solutions to be explored by the Maicholl Gutierrez, Pontificia PLS algorithm at each iteration, leading to a Universidad Católica de Valparaiso, faster convergence with a minimum Chile, [email protected] impairment on the final treatment plans found by our approach. Gustavo Gatica, Universidad Andres Bello, Chile, [email protected] References Guillermo Cabrera-Guerrero, Carolina Lagos, Pontificia Universidad Andrew J. Mason, Andrea Raith, Matthias Catolica de Valparaíso, Chile, Ehrgott, 2018. "Pareto local search [email protected] algorithms for the multi-objective beam angle optimisation problem," Journal of

141

Heuristics, Springer, vol. 24(2), pages 205- oncological clinics. We made use of 238, April. evolutionary singleobjective and multiobjective optimization methods. Our 3. Dimensionality Reduction for results are comparable to those obtained Multiobjective Optimization via from a commercial radiotherapy planning Thermal Maps – The Case of system. The choice of evolutionary Intensity Modulated Radiotherapy optimization was motivated by our want to easily incorporate into the treatment Przemysław Juszczuk, University of planning information about various Economics, Katowice, Poland, radiobiological effects modelled as nonlinear [email protected] objective functions (e.g., tumor control Ignacy Kaliszewski, Polish Academy of probability). As yet, commercial treatment Sciences, Poland, planning systems do not account for such [email protected] effects. *Janusz Miroforidis, Polish Academy of 4. Evaluating the Quality of Sciences, Poland, Radiotherapy Treatment Plans for [email protected] Prostate Cancer Dmitry Podkopaev, Polish Academy of Sciences, Poland, *Matthias Ehrgott, Lancaster University, [email protected] United Kingdom, Robert Szmurło, Warsaw University of [email protected] Technology, Poland, Emma Stubington, Lancaster University, [email protected] United Kingdom, Anna Zawadzka, The Maria Skłodowska- [email protected] Curie Institute of Oncology, Poland, Omid Nohadani, Northwestern [email protected] University, United States, [email protected] Large-scale multiobjective nonlinear Glyn Shentall, Royal Prestom Hospital, optimization problems can be intractable by any methods, either exact or evolutionary. In United Kingdom, such cases, two major approaches are [email protected] available: decomposition and dimension External beam radiation therapy is a (size) reduction. In multiobjective common treatment method for cancer. optimization problems arising in Intensity Radiotherapy is planned with the aim to Modulated Radiotherapy (IMRT), problem achieve conflicting goals: while a sufficiently sizes quickly explode when discretization high dose of radiation is necessary for grids become fine. By necessity, radiation tumour control, a low dose of radiation is plans are Pareto optimized on coarse grids desirable to avoid complications in normal, which can lower the quality of radiation healthy, tissue. These goals are encoded in plans. We present an approach in which an clinical protocols and a plan that does not IMRT case is defined on a fine grid and the meet the criteria set out in the protocol may resulting large-scale multiobjective have to be re-optimised using a trial and optimization problem is run for a number of error process. To support the planning iterations to find a good feasible solution. process, it is therefore important to evaluate Then, variables are aggregated on the base the quality of treatment plans in order to of some heuristic which uses for the input recognise plans that will benefit from such thermal maps intensities, as defined by the re-optimisation and distinguish them from feasible solution, of rays emanating from the those for which this is unlikely to be the case. radiation source. Finally, the aggregated In this talk we present a case study of problem is run to optimality. Our approach evaluating the quality of prostate cancer allows to account for the specificity of treatment plans based on data collected individual patient organ-tumor structures from Rosemere Cancer Centre at the Royal which are reflected, roughly, in the feasible Preston Hospital in the UK. We use Principal solution to the large-scale multiobjective Component Analysis to select the most optimization problem. We experimented with relevant data. We then apply Data real data provided by one of the Polish Envelopment Analysis to assess the quality 142 of individual plans. Each plan is compared 1. Convex Cones and Column against the entire set of plans to identify Generation, Part I: Perfect those that could realistically be improved. We further enhance this procedure with Information simulation techniques to account for *Ozgur Ozpeynirci, Izmir University of uncertainties in the data for treatment plans. Economics, Turkey, The proposed approach to plan evaluation [email protected] provides a tool to support radiotherapy Selin Özpeynirci, Izmir University of treatment planners in their task to determine the best possible radiotherapy treatment for Economics, Turkey, cancer patients. [email protected] THU-2-P We work on the multiple criteria selection problem that aims to find the most preferred Plenary Session 4 alternative among a set of known alternatives evaluated on multiple criteria. Thursday 11:10-12:10 - Room: Mustafa We assume the decision maker (DM) has a Kemal Amfisi quasi-concave value function that represents Chair: Ilker Topcu his/her preferences. The DM provides pairwise comparisons of alternatives and Voting with Intensity of Preferences this information is used to generate convex *Luis Vargas, University of Pittsburgh, cones (Korhonen et al., 1984). The generated cones eliminate inferior United States, [email protected] alternatives and the most preferred In this paper we develop a method based on alternative is identified after enough pairwise the idea of pairwise voting to rank projects or comparison information is gathered. candidates and incorporate in the ranking Generally, an interactive approach is used to process how strongly the referees/voters obtain the preference information of the DM feel about the comparisons they make. (Köksalan et al., 1984; Köksalan, 1989; Voting is a modified form of ranking and all Taner and Köksalan, 1991; Köksalan and the votes are equally important. However, Taner, 1992; Özpeynirci et al., 2017). The there are situations like voting in which the aim of interactive algorithms is to detect the votes are not just ordinal, but each voter most preferred alternative by performing as expresses an intensity of preference for the few pairwise comparisons as possible. In different candidates, e.g., ranking projects this study, we aim to compute the minimum for funding. We show that our method yields number of pairwise comparisons required to the same results as ordinal voting in large identify the most preferred alternative. For populations when the intensity of this purpose, we assume that the value preferences becomes extreme. Voting with function of the DM is known. If all convex intensity of preferences does not violate cones are generated, then it is possible to democracy but soften the stand of voters and find optimal pairwise questions and convex allows for consideration of the diversity of cones to eliminate all alternatives but the issues involved in voting. most preferred one. Özpeynirci et al. (2017) generate all single and double cones (cones THU-3-U that require two pairwise comparisons) and Business Meeting find an upper bound for the objective function value. It is not practical to generate Thursday 13:10-14:10 - Room: Mustafa all cones since Köksalan et al. (1984) show Kemal Amfisi that the number of convex cones grows exponentially with the number of Chair: Murat Köksalan alternatives. We develop an approach that THU-4-A utilizes column generation for generating cones under the perfect information Invited Session: Hybrid Approaches in assumption. We then, embed this approach MCDM in a branch and bound framework to guarantee the optimality. We conduct Thursday 14:20-16:00 - Room: Galata Tower computational experiments on different Chairs: Esra Karasakal, Orhan Karasakal problem instances and report the results. 143

References Korhonen P., Wallenius J., Zionts 3. Convex Cones and Column S., Solving the discrete multiple criteria Generation, Part II: Interactive problem using convex cones. Management Science, 1984, 30, 1336–1345. Köksalan Algorithm M. Identifying and ranking a most preferred *Selin Özpeynirci, Izmir University of subset of alternatives in the presence of Economics, Turkey, multiple criteria. Naval Research Logistics [email protected] 1989;36:359-372. Köksalan M., Karwan M., Ozgur Ozpeynirci, Izmir University of Zionts S., An improved method for solving multiple criteria problems involving discrete Economics, Turkey, alternatives. IEEE Transactions Systems Man [email protected] Cybernetics, 1984;14:24–34. Köksalan M., Multiple criteria selection problem aims to Taner O., An approach for finding the most find the most preferred alternative among a preferred alternative in the presence of set of known alternatives evaluated on multiple criteria. European Journal of multiple criteria. We assume the decision Operational Research 1992;60:52–60. maker (DM) has a quasi-concave value Taner O., Köksalan M., Experiments and an function that represents his/her improved method for solving the discrete preferences. The DM provides pairwise alternative multiple-criteria problem. Journal comparisons of alternatives and we use this of the Operational Research Society information to generate convex cones 1991;42:383–91. Özpeynirci Ö., Özpeynirci (Korhonen et al., 1984). We eliminate the S., Kaya A., An interactive approach for alternatives that are dominated by these multiple criteria selection problem. cones and identify the most preferred Computers & Operations Research 2017; alternative after certain number of iterations. 78: 154-162. In the literature, there are several studies 2. A Mathematical Programming that consider convex cones and use interactive approaches to obtain the Evaluation Approach for Multiple preference information of the DM (Köksalan Criteria Sorting Problems et al., 1984; Köksalan, 1989; Taner and *Merve Civelek, ASELSAN, Turkey, Köksalan, 1991; Köksalan and Taner, 1992; Özpeynirci et al., 2017). The aim of [email protected] interactive algorithms is to detect the most Esra Karasakal, Middle East Technical preferred alternative by performing as few University, Turkey, pairwise comparisons as possible. In this [email protected] study, we develop an interactive algorithm Multiple criteria sorting problem is to assign that uses convex cones and column alternatives, evaluated according to multiple generation. We get some initial information criteria, into predefined preference ordered from the DM and estimate the value function. classes. In this study, a new distance metric Throughout the iterations, the interactive based sorting method is developed to solve algorithm seeks for the cones that can be multiple criteria sorting problems. The aim of generated with the currently available the proposed method is to assign each information including transitivity using the alternative to one class or a set of possible column generation approach. After the adjacent classes. In the proposed method, alternatives dominated by the generated centroids of the classes are estimated using cones are eliminated, the algorithm sample preference set provided by the estimates the benefit of asking a specific decision maker. Distance to the centroids pairwise comparison to the DM. This benefit are used as a criteria aggregation function. A is calculated based on the dual prices mathematical model is formulated to provided by the mathematical model that determine the weights of the distance metric. aims to minimize the total number of questions with the currently available Assignment is performed according to the weighted distance of each alternative to information. The pair of alternatives that is each class’ centroids. The proposed method estimated to eliminate the largest number of is applied to different data sets and its alternatives is presented to the DM. performance is compared with other Necessary updates on the mathematical methods in literature. model and the remaining alternative set are done. The procedure continues until there is 144 only one alternative left. We conduct classification phase of ATR and develop a computational tests and compare the results novel multiple criteria classification method with the minimum number of questions based on modified Dempster Shafer data found by convex cones and column fusion algorithm. Ensemble of classifiers generation under perfect information. are used as a classification algorithm. They References Korhonen P., Wallenius J., Zionts are treated as the state of the art technology S., Solving the discrete multiple criteria for classification in which each single problem using convex cones. Management classifier is trained separately, and then the Science, 1984, 30, 1336–1345. Köksalan results of them are combined through M., Identifying and ranking a most preferred several fusion algorithms. Support vector subset of alternatives in the presence of machine and neural network are employed multiple criteria. Naval Research Logistics as probabilistic classifiers in ensemble. Each 1989;36:359-372. Köksalan M., Karwan M., non-imaginary dataset coming from multiple Zionts S., An improved method for solving heterogeneous sensors is classified by both multiple criteria problems involving discrete of the classifiers in the ensemble, and the alternatives. IEEE Transactions Systems Man classification result that has higher accuracy Cybernetics, 1984;14:24–34. Köksalan M., ratio is chosen for each of the sensor Taner O., An approach for finding the most dataset. After getting probabilistic preferred alternative in the presence of classification of targets by different sensors, multiple criteria. European Journal of modified Dempster Shafer data fusion Operational Research 1992;60:52–60. algorithm is used to combine the sensors’ Taner O., Köksalan M., Experiments and an results to reach the final classification of the improved method for solving the discrete targets. In this talk, a number of alternative multiple-criteria problem. Journal classification algorithms are compared with of the Operational Research Society the proposed algorithm and the results will 1991;42:383–91. Özpeynirci Ö., Özpeynirci be discussed. S., Kaya A., An interactive approach for multiple criteria selection problem. THU-4-B Computers and Operations Research 2017; Contributed Session: Business 78: 154-162. Applications of AHP/ANP 4. Multiple Criteria Target Thursday 14:20-16:00 - Room: Maiden's Classification Using Heterogeneous Tower Sensor Data Chair: Ilker Topcu *Orhan Karasakal, Çankaya University, 1. A Geometric Standard Deviation Turkey, [email protected] Based Soft Consensus Model in Bengü Atıcı, ASELSAN, Turkey, [email protected] Analytic Hierarchy Process Esra Karasakal, Middle East Technical *Petra Groselj, University of Ljubljana, University, Turkey, Slovenia, [email protected] [email protected] Lidija Zadnik Stirn, University of Radar systems have important roles in both Ljubljana, Biotechnical Faculty, military and civilian applications. As the Slovenia, [email protected] capabilities increase in terms of range, Sophisticated real world problems can be sensitivity and the number of tracks to be successfully solved by group multi-criteria handled, the requirement for automatic decision-making approaches that are target recognition (ATR) increase. ATR recognized as reliable and effective. To systems are used as decision support efficiently tackle such problems a group of systems to classify the potential targets in decision makers with different opinions, military applications. These systems are knowledge and experiences should be composed of four phases, which are included in the decision-making process. selection of sensors, preprocessing of the Analytic hierarchy process (AHP) is one of the radar data, feature extraction and selection, well-known approaches for handling group and processing of features to classify the multi-criteria decision making problems. potential target. In this study, we focus on the Aggregating individual judgments into group

145 judgments is a challenging topic in AHP and Anjan Kumar Swain, Indian Institute of needs to be researched more in depth. The Management Kozhikode, India, best known aggregating method is the [email protected] weighted geometric mean method, but the decision makers are usually not satisfied Price is the only “P” of marketing mix that with this technical solution. A certain level of generates revenue and therefore how firms agreement among decision makers should set and adjust their prices in order to increases the effectiveness of the decisions. maximize profitability is very important. Of course, consensus which would represent During last few decades, many quantitative a unanimous agreement of all decision techniques were developed to dynamically makers is an ideal solution; however it is adjust prices so that the right prices are in difficult to achieve such ideal solution in the right place at the right time, through the heterogeneous groups included in decision- right channel and to the right customer. making process of real world problems. It is However, most of these quantitative already known that consensus and soft techniques are based upon various inherent consensus reaching models can lead to assumptions and use limited number of more unified and effective decisions and parameters and variables. In addition, these more satisfied decision makers. A consensus quantitative techniques often use precise process involves several consensus rounds historical data but provide little scope to and in each round the degree of consensus incorporate the preferences of the decision is measured. The level of agreement in a maker (DM) in the process of making pricing group of decision makers can be measured related decisions. Studies have shown that by the distance between the priority vectors, the judgments are always inferior to by compatibility index or any other forecasting models because forecasting consensus measure. We propose a new soft models use precise historical data. However, consensus model based on a geometric the quantitative model lack in capturing all standard deviation as a measure of the relevant realities of the environment. In closeness of group opinions. Geometric certain situations, the judgments of the standard deviation describes the variability decision maker can give superior results. of a set of numbers around their geometric Further, quantitative models are well suited mean. One of the important aspects of the to estimate an exact price offering but consensus reaching process is that decision judgments based models can very well be makers can decide whether they are willing used to estimate price brackets and to adapt their judgments. In AHP consensus positioning. Setting an exact price offering reaching models which are found in literature often depends upon the price brackets and it is assumed that once a decision maker positioning. This is because the ultimate rejects to adapt his judgments, he cannot price figure should always align with the change his mind in the future iterations. We marketing/pricing strategy of the firm. found this fact to be a deficiency and in our Deciding marketing/pricing strategy of the proposed model the decision maker can firm require processing various quantitative change his mind. At the end, we apply the and qualitative indicators of the market as model to an environmental problem about well as of the firm. This requires the development of a mountainous area. understanding of the market through the Several stakeholders are included in the expertise, knowledge, and preference of the decision making process with economic, pricing expert or the decision maker. environmental and social criteria and However, there is no decision framework objectives. The purpose of the application is available in the literature to take a decision to compare the new proposed consensus on the pricing strategy of a firm using the model with other existing models. preference or the judgment of the pricing expert or the marketing manager. Using the 2. Strategic Pricing Decision using findings of Rao and Kartono (2009), this Analytic Hierarchy Process research work proposes a decision hierarchy that can be used to take a decision on the *Ram Dhurkari, IFMR Graduate School pricing strategy of a firm. The decision of Business Sricity, India, hierarchy begins with the overall objective of [email protected] setting or selecting a best pricing strategy for the new/existing product or services of the

146 firm. Down further, the decision hierarchy should bring the company closer to achieving contains two criteria, six sub-criteria, and the set goals. Besides the knowledge, skills three possible strategic alternatives as and experience it possesses, the shown in the figure 1. The two broad criteria management should also emphasize the considered here are the objectives of the quantitative analysis. Good measurement firm and the market determinants. To select enables results to be obtained, which should a best pricing strategy for the firm, the serve as a recommendation for the decision maker will define the relative management’s decision-making process. importance between the objectives of the The aim of the research in this article is to firm and the market determinants. Each of examine how much the Management these two criteria is measured with the help Science (MS) models and methods are of three sub-criteria. The decision maker will applied in companies in a certain developing then define the relative importance of these country, i.e. the Republic of North three sub-criteria with respect to the criteria Macedonia, what effects have been under which they are placed in the decision achieved by their application, and which hierarchy. This is required because a firm skills MS consultants should have in regards may have several objectives to achieve but to being engaged in solving the problems all will have different priorities with respect to that companies face. The survey was the overall objective of the firm. Similarly, conducted via a questionnaire given to several market determinants are required to senior managers of state and private be assessed and their relative priorities are companies in the Republic of North to be defined in terms of their importance in Macedonia. The total number of received shaping the pricing strategy of the firm. questionnaires is 236; out of which 17 are Leaves of the decision hierarchy are the incomplete, so 219 have been analyzed. Out three possible alternatives. The decision of these 219 questionnaires, 68 were filled maker will then define the relative out by senior managers of state companies, importance of the three alternatives while 151 of them came from private separately and independently with respect to companies. Most of the companies (154) are each of the sub-criteria placed immediately located in the capital of the country, Skopje. above the alternatives in the decision In terms of the industry, most state-owned hierarchy. Since the assumptions of the companies (13 of each) fall under public problem structure closely match with the administration and defense: compulsory axiomatic assumptions of the Analytic social security and health and social care Hierarchy Process (AHP), a popular multi- activities, while most of the private criteria decision making method, this companies (40) are from the manufacturing research also demonstrates how the AHP industry. Considering gender, 113 are males method can be used to solve the strategic and 106 are females, while the average age pricing decision problem. is 43. Regarding the level of education, the majority of the respondents (143) have 3. Skills of A Management Science gained a University education. Most of the Consultant: AHP Model through the respondents (151) answered that there is a Perspective of Senior Managers in a need for MS support, and that in the future Developing Country they plan to hire an MS consultant, while 68 respondents said that they do not need such *Violeta Cvetkoska, Ss.Cyril and support nor engagement from an MS Methodius University in Skopje, North consultant. The skills that an MS consultant Macedonia, should have are grouped into two categories: [email protected] fundamental and ancillary practical skills. On the basis of the estimated average grade of In order for companies to survive and importance of the fundamental skills, the MS develop in this dynamic and complex world, consultant able to view the problem was it is necessary for them to adapt to changes given the highest average, followed by: being and to invest in employee education. able to connect the results to the real Efficient management is not only about situation, structure a problem, analyze the achieving results, but also gaining the trust model, relate to the client, model the and respect of employees. The management problem, identify opportunities for analysis in of companies with the decisions it makes a creative manner, and comprehend the

147 social geography of the client body. Ancillary company; some of them should be selected practical skills are divided into five fields: because of time, workforce, and monetary marketing, sales, formal communication and constraints. This study proposes an reporting, skills in interacting with the client integrated multi criteria decision making aid and facilitating, and computing; while the for the executives of the companies in average grade of importance for each skill in private sector in their selection decisions of these fields is estimated and analyzed. The research and development projects. After emphasis in this article is to develop a defining evaluation criteria through literature structure of a multi-criteria decision model survey and experts’ opinion, the relations (AHP) for managerial decision-making among them were revealed using cognitive regarding hiring an MS consultant based on mapping approach. Cognitive maps are their fundamental and ancillary practical cause-effect networks, with nodes skills from the point of view of the senior representing concepts articulated by managers in both private and state individuals, and directional linkages companies in the Republic of North capturing causal dependencies. Then, Macedonia. Only the skills with an average analytic network process (ANP) method was grade of importance that is not below four utilized to determine the importance of will serve as input in the AHP model. The criteria. ANP is a methodology that allows constructed multi-criteria decision model is groups or individuals to deal with the presented and it will be tested in both state interconnections (dependence and and private companies in the country. With feedback) between factors of complex this kind of research we want to introduce structure in decision making process. As a the management in our country with the further step, in a case study, the global benefits of using MCDM methods and scores of candidate projects were computed models in their companies, especially for the using simple additive weighting (SAW) purpose of making better decisions that will method. A global score in the SAW is lead to successful operating and growth in obtained by adding normalized contributions future. Key words: MCDM, AHP model, of alternatives with respect to each attribute. developing country, MS consultant, skills Finally, goal programming was used to select best combination of candidate projects under various constraints regarding to company strategies and decision maker preferences. Objective function of the 4. An Integrated Multi Criteria optimization model is minimizing deviations Decision Making Method for R&D of conflicting goals. Goals of the model consist of (i) expected net present value of Project Selection projects, (ii) global scores of candidate *Gizem Filiz Türkmen, Istanbul projects and (iii) funding potential of each Technical University, Turkey, project. Constraints include dependencies [email protected] between projects, limits related to expected Ilker Topcu, Istanbul Technical contributions of projects and source constraints like time, budget and human University, Turkey, [email protected] resource. Research and development activities can be considered as putting forward and creating THU-4-C new products or services as well as Invited Session: MCDM in Solving Urban improving existing products or services by Problems using Big Data using technology, knowledge and experience. Research and development Thursday 14:20-16:00 - Room: Dolmabahçe activities serve significant roles for the Palace countries and societies as these activities Chairs: Sehnaz Cenani may lead inventions, innovations, technological developments, economic 1. Anchoring in Two-Criteria Trade- growth, and better living conditions. Although Offs between Carbon Dioxide there are several possible research and Emission and Travel Time development projects, those can be executed by an institution, an agency or a

148

*Fredrik Bökman, University of Gävle, information, the anchoring effect is even Sweden, [email protected] more pronounced. Although the experiment Hanna Andersson, University of Gävle, shows no significant main effect of the Sweden, [email protected] additional normative message, there is an Ulla Ahonen-Jonnarth, University of interaction between anchoring and the given information. The results of our experiments Gävle, Sweden, [email protected] are important in multiple criteria decision- Anchoring is one of the most common making since similar kinds of trade-offs and cognitive biases people may be influenced by matching procedures are common in when making decisions, estimates, and practical elicitation procedures. E.g., the predictions. Here we will present the effects situation corresponds to making trade-offs in of anchoring and framing in a two-criteria the even swap method. trade-off setting. The context is a personal decision on reducing carbon dioxide 2. Understanding Divisions of a City emission from air travel, which is currently a by Evaluating Location-Based Social hot topic in Sweden. Participants in a Media Data between-subjects experiment were asked to make a trade-off, matching a reduction in *Kutay Yunculer, Istanbul Technical CO2 emission from 99 kg to 22 kg to an University, Turkey, [email protected] increase in travel time from 1 hour to a value For municipalities, urban designers, and they chose themselves. However, the architects, defining the divisions of a participants were anchored by first having to metropolis is an essential part of developing consider if they would be willing to increase urban strategies. However, only using city the travel time to more than 2 hours (low districts’ provincial borders is not illustrating anchor condition) or 6 hours (high anchor the whole picture because of the two main condition). Robust data analysis methods deficiencies. First, in most of metropoles, a were used for hypothesis testing. As single district contains multiple partitions expected, the answers from participants in with different specialities and mostly these the group with a high anchor were partitions are not defined by borders of the significantly higher than answers from the district which they belong. For instance, if the group with a low anchor. Since reducing urban-scale network of Istanbul is explored private air travel is often proposed as a by the method which is explained in the component of reducing one’s carbon second paragraph, one can see that footprint, we supposed that many historical areas of Istanbul named Eminönü participants in the study should be willing to and Sultanahmet which are actually in Fatih extend the travel time in order to reduce the district are in the same cluster with another CO2 emission. However, it may be difficult to district named Beyoğlu instead of Fatih. realize how much this reduction from 99 to Secondly, in rapidly changing cities such as 22 kg CO2 is worth. In order to put this Istanbul, characteristics of these districts are amount in context, two other groups of changing continuously. At this phase, participants in the study received an research projects such as Livehoods [1] are additional piece of information in the context remarkable by showing us, how promising it description: “The Swedish Environmental can be to use data gained by location-based Protection Agency recommends that each of social media networks’ APIs. Despite missing us should not use more than two tons of some necessary information due to greenhouse gases per person and year, limitations and constraints (i.e. limited which amounts to a maximum of 38 kg CO2 sample of check-ins, not implementing per week on average.” This additional centrality algorithms to the study), Livehoods information thus puts the emissions of 99 emphasised the possibility of picturing urban and 22 kg CO2 in context, being in the same divisions by using location-based social order of magnitude as a maximum weekly media data and clustering algorithms. This recommended use. Further, this additional research is aiming to adapt approaches of information conveys a normative message. It previous research such as Livehoods to may also be seen as a case of issue framing, current versions of location-based social where we have emphasized the issue of a media networks’ APIs. More importantly personal responsibility for reducing combining these approaches with some greenhouse gases. With the additional more recent research, which were mainly 149 focusing on analysing and determining *Begum Moralioglu, Istanbul Technical significancies of venues in a city by using University, Turkey, centrality algorithms such as PageRank. With [email protected] this methodological motivation, this study Sehnaz Cenani, Istanbul Technical targets contributing urban decision making University, Turkey, processes in two different ways. First and most important goal is developing and [email protected] publishing a Python library which collects Gulen Cagdas, Istanbul Technical urban-scale data from location-based social University, Turkey, [email protected] media networks’ APIs such as Foursquare’s A smart and sustainable city should be an Places API then converts this data into a innovative city that uses information and GEXF graph file which is proposed by communication technologies (ICT) to developers of Gephi. Second goal is directly improve the quality of life via its operations. related with the first one by aiming to be a They need to be planned, managed and proof-of-concept of it. To achieve this goal, an regulated by open data collected through urban-scale graph of Istanbul and Stuttgart different data sources to provide efficient will be produced and stored as GEXF files by services. It can be said that one of the using the Python library which is developed essential services of a city is transportation. as part of this study, then these graphs will In smart cities, intelligent transportation be evaluated by analysis tools such as systems help to solve problems such as PageRank and Modularity algorithms which traffic congestion or the amount of fuel spent are implemented to Gephi, an open-source in traffic by providing communication data visualization and exploration software. between vehicles and devices that build the Main motivation while choosing these two whole transportation network. In order to cities was their networks’ distinct achieve the success of intelligent characteristics from other cities and from transportation systems, transportation each other. To be more precise, when methods should be planned dynamically Istanbul’s network is explored in Gephi, an according to the collected data and the urban-scale graph which consists of 28 requirements of the city's transportation interwoven clusters will be seen. On the network. Then the necessary decisions contrary, when a researcher analyses should be made with the help of all the data Stuttgart’s urban-scale network by the same and information to improve the technique will see discrete clusters. transportation system. Today in 2019 as we Furthermore Stuttgart has a strong move from the information age to the digital connections with neighbor cities especially age, the possibilities of data usage may with Ludwigsburg, conversely intercity reshape the decision support systems. Social connections cannot be noticed in the case of media is one of the new tools that collects Istanbul. Last but not least, the result of the valuable information like geolocation data to evaluation process will be published as an be used in various decision support systems interactive website which will let researchers, for different industries. The location tags that professionals and citizens to create their own are used in location-based social networks reports about Istanbul and Stuttgart. Thus, like Instagram and Twitter can have an this website aims to bring processed impact on the studies related to urban information such as clusters in these cities, design and planning. In this study, we distribution of venue types in these clusters, present an electric scooter deployment average centrality of venues in different model to reduce traffic congestion and travel clusters, and average centrality of venue time with the use of real-time data from types in the whole city as well as in different social media in urban areas with heavy clusters, together with the essence of pedestrian and vehicle traffic. According to exploratory data-visualization concept. [1] the Turkish Statistical Institute, in 2018 http://livehoods.org/ more than 15 million people live in Istanbul 3. A Decision Support System to and the number of motor vehicles is more than 4 million. Therefore, Istanbul is chosen Provide Real-Time Alternatives for as the case study for the proposed model, Placement of E-Scooters in a Smart since transportation is one of the most City: A Case Study of Istanbul critical problems of the city. Sustainable urban mobility is a feasible solution for 150 transportation-based problems. Electric Sehnaz Cenani, Istanbul Technical vehicles are cost-effective and sustainable University, Turkey, transportation methods. Among them, e- [email protected] scooters can easily fit into small and narrow Gulen Cagdas, Istanbul Technical streets and reduce parking problems in University, Turkey, [email protected] cities. It is also an exciting transportation method especially preferred by young Disasters have devastating effects on people. In the scope of this study, a survey socioeconomic, built environment and is conducted to estimate the e-scooter infrastructure; also create a volatile, potential usage in Istanbul. The main criteria uncertain, complex and ambiguous for the locations of e-scooters are collected processes which are hard to manage. through online questionnaires and Therefore, developing a proper disaster information available from the e-scooter management strategy is a crucial need, companies around the world. Following that, especially for the governments. Also, collected criteria are processed in multi- disasters need excessive efforts on logistical criteria decision analysis. We use the Analytic and organizational aspects of the affected Hierarchy Process (AHP) method to weight country. So, using traditional methods while and rank these criteria for sharing location developing this strategy may not be efficient; selection. This paper proposes the use of the therefore, it is crucial that the implementers AHP method for application in the smart and of disaster management should adopt sustainable city context. Thus, a decision contemporary, rapid, precise and effective hierarchy is built for setting the priorities for methods in the process of successful e-scooter mobility in an urban environment. humanitarian aid and disaster relief. In The Super Decisions, which is an open- general, there are three basic phases of source decision support software that disaster management, which are: (1) implements the AHP and ANP, is used to preparedness: which consists the activities evaluate decision processes in this study. of pre-disaster period, (2) response: which Then, the methodology is used to develop a consists the activities of disaster period and prototype for an application that collects (3) recovery: which consists the activities of pedestrian and vehicle traffic information post-disaster period. In pre-disaster period from different districts and location tags from mitigation and preparedness activities are social networks in real time. All traffic held in a strategic planning approach. In information is collected through transactions disaster period, response activities take of Google Maps users and cross-checked place, so agile principles should be taken with the survey results. The short-distance into account in this period as short term routes with the high pedestrian and vehicle project management, flexibility and traffic in the city are categorized in the nimbleness are the most important subjects. decision process. In that way, the location In after-disaster period, recovery activities alternatives to place e-scooters will be replace the premises. Preparedness phase changing according to data coming from needs proactive approaches, as the other traffic information applications and social ones need reactive approaches. One of the media. We assume that this system will offer hardest things in disaster management is the best locations for e-scooter placement the trade-off between cost and and use. Additionally, the proposed approach responsiveness; also, it’s a hard and fragile can reduce commuting time, while process that needs quick actions. To achieve contributing to the reduction of effective, efficient and timely responses environmental footprint. during and right after disasters; strategic planning should be done long before the 4. Selection of Temporary Shelter happening. Selecting appropriate shelter Sites and Optimising Evacuation area and routing for the safest evacuation is Routes as a Disaster Management the key to this planning process, especially Strategy: A Case Study for the for the densely populated cities. This paper Expected Istanbul Earthquake mainly focuses on the preparedness phase of the disaster management as offering a *Sezer Savaş, Istanbul Technical decision-making tool to support the University, Turkey, [email protected] government for selection of temporary shelter sites and optimizing evacuation

151 routes of disaster victims. This selection and *F. Antonio Medrano, Texas A&M optimization process needs multi-disciplinary University-Corpus Christi, United States, approach involving many different [email protected] knowledge areas including management, seismology, environment, sociology, The vertex p-Center problem consists of planning, architecture, law, transportation, locating p facilities that cover all n demands engineering and so on. As a case study, the in order to minimize the maximum distance expected Istanbul earthquake is chosen in between a demand and the facility that terms of what-if scenario of a very disastrous covers it. In other words, it is equivalent to earthquake to demonstrate the finding the minimum coverage radius r and implementation of the proposed model. In locations of p facilities capable of covering all this paper, we analyze the principles of site demands. Originally formulated by Hakimi selection of temporary shelters from the (1965), this problem is NP-Hard, and the relevant literature and prioritize the standard formulation has proven to be important criteria and sub-criteria. The particularly challenging for IP solvers on large analytic hierarchy process (AHP) is used to problems. Thus, most solution approaches find criteria weights. As a multi-criteria use heuristics or relaxations (Minieka 1970, decision problem, we formulate the selection Tansel et al. 1983, Mladenović et al. 2003, of the most appropriate shelter sites in terms Elloumi et al. 2004). The complete p-Center of location, capacity, resources, allocation of problem extends the formulation to solve the victims and evacuation route principles. for all p-values from 1 to N, where N is the Then, we compare them with the current number of nodes to cover. This provides a areas defined by the local government. As a complete coverage trade-off curve between result, improvements were achieved on number of facilities and coverage radius. comprehensiveness among population, This talk proposes an approach for solving averaged evacuation distance traveled by the Complete Vertex p-Center Problem using the victims and reaching times of the an iterative location set covering approach affected victims to the temporary shelter with selecting only r-values that have the sites when compared to the current ones. potential to be exact solutions. Experiments This study offers a tool for policy-makers in on a variety of data sets demonstrate that terms of effective decision-making support this new covering trade-off method is faster on shelter location planning. The aim of this than approaches using traditional p-Center study is to better understand the criteria formulations, and in certain instances can be related to the post-disaster shelter location further sped up with brute force planning and to design a decision support combinatorics. system for policy-makers in the smart city framework. For the further studies, it is 2. Multi-Objective Combinatorial proposed to analyze the described Optimization of Coupled Systems methodology and use the proposed system *Antoine Kerbérénès, Université Paris on previous and future disaster situations Dauphine, France, [email protected] occurred in different places of the globe to observe whether it is performing well in Daniel Vanderpooten, Université Paris different real-life scenarios. Dauphine, LAMSADE, France, [email protected] THU-4-E Jean-Michel Vanpeperstraete, Naval Contributed Session: Practical Group, France, jean- Approaches to Challenging Problems in michel.vanpeperstraete@naval- Multiobjective Optimization group.com We take interest in the optimal functioning of Thursday 14:20-16:00 - Room: Hagia Sophia systems which can usefully be seen as Chair: Ralph Steuer composed of distinct subsystems. This may be the case because most constraints 1. A Set Covering Trade-Off to Solve defining the global problem are local to the Complete Vertex p-Center particular subsystems, while these Problem subsystems still are linked together by other constraints. Such systems have been called

152 coupled, complex or even interwoven. They However these authors do not consider the can model multi-site production problems, global vector-valued objective function as an resource allocation problems, multi-agent objective-wise aggregation of local objective task planning and other such coordination functions along subsystems, rather they problems. The problems we consider are project the objective space of each sub- both globally multi-objective, and such that problem into a subspace of the global each objective function is additively problem's objective space. Finally, there is, to separable along the subsets of variables our knowledge, no literature dedicated to characterizing the subsystems: it is the sum coupled systems with integer variables, in of terms which relate to subsystem which subproblems can themselves be NP- problems. Thus, each restriction of the hard. After having defined notions which are problem to variables characterizing a key to the study of the multi-objective subsystem defines a local multi-objective combinatorial optimization of coupled problem for the same number of objectives systems, we will present a reference as the global problem. Such a definition problem, which is of particular interest to us suggests a decomposition of the problem because both the global problem and its into subproblems, and a coordinated subproblems can be solved using multi- resolution of the subproblems. The weaker objective dynamic programming, which the coupling between the subsystems, the generalizes easily beyond two criteria. We more relevant decomposition proves to be, will also propose a multi-objective Branch & for example relaxing the coupling Bound approach to the resolution of coupled constraints. Such strategies have been problems, which uses decomposition to extensively studied in the single-objective quickly obtain lower and upper bounds, and case and with continuous variables. It should a branching scheme specifically designed to also be noted that the multi-objective progressively decouple the problem and versions of problems with "block angular" reduce it to a significantly easier version of structures are a particular case of the class itself. In particular, from a certain depth of of problems under consideration. Our work the search tree, the problem will be fully differs from two main other approaches to decoupled and decomposition will be the optimization of coupled systems. One is available to speed up the resolution using, multidisciplinary optimization, in which, once more, dynamic programming. although global problems may be formulated as multi-objective, each subsystem problem 3. Rationalizable Strategies for is single-objective. Multidisciplinary Competing TSP with Multiple-Criteria problems are also endowed with a *Erella Eisenstadt-Matalon, ORT hierarchical structure of the subsystems, and thus benefit from natural coordination Braude Colleage, Israel, schemes. In our case, the absence of such [email protected] properties is an obstacle to the Amiram Moshaiov, Tel-Aviv University, decomposition of the global problem into a Israel, [email protected] collection of parameterized subproblems, The presented study concerns a novel such that the optimal solution for a method for solving Multi-Objective Games subproblem and a parameter value would (MOGs), which are also known as vector- provide a useful information for the pursuit payoff games, multi-payoff games and multi- of resolution - if only because a subproblem criteria games. The considered MOGs are has not one but many incomparable optima. non-cooperative, two-persons, zero-sum In addition, theoretical results suggest that games with pure strategies, in which each combining local optimal solutions in order to player has self-conflicting objectives and get optimal solutions to the global problem none of the players has a-priori objective cannot be done as straightforwardly in the preferences. Yet, each player knows all the case of Pareto efficiency as it can be done in available strategies of the opponent and all the case of single-objective optimality. The the payoff vectors that result from all second approach we differ from, stems from possible interactions between their own and recent work by Dietz & al. (2018), who their opponent’s strategies. The main proposed a framework for complex systems assumption is that not only that each player multi-objective optimization in which is rational but also that each player knows it. subproblems are themselves multi-objective.

153

Namely, common knowledge of rationality is 4. Reducing Times for All Efficient assumed. The considered MOGs involve Extreme Point Computation in incomplete information that induces partial order to the outcomes of the game. MOLPs with Larger Numbers of Traditionally, such MOGs have been solved Objectives either by an equilibrium solution concept or Ralph Steuer, University of Georgia, by a MiniMax solution concept. Yet, existing United States, [email protected] studies suffer from not considering performance trade-offs. Namely, they *Craig Piercy, University of Georgia, assume that the players have no preference United States, [email protected] of objectives when selecting a strategy. In The paper addresses unresolved issues with contrast, this study is based on the assertion regard to the problem of partitioning the that in general decision-makers should take criterion cone of a multiple objective linear into account performance trade-offs when program (MOLP) into sub-cones for the making a decision. In view of the purpose of solving an MOLP in parallel to aforementioned assertion and the state-of- achieve dramatically reduced times in all the-art, a novel solution concept to MOGs efficient extreme point computation. One of has been recently suggested by the authors. the issues discussed is the declining In contrast, following the solution concept of effectiveness of the approach as the number rationalizability, here the solution concept is of objectives increases. Another is the task revised into mutual-rationalizability that of balancing the size of the sub-cones in involves the assumption of common accordance with the size of their spherical knowledge of rationality. The proposed caps. Numerous graphs are shown and approach, for solving MOGs, involves two- some interesting computational experience stages. In the first stage, a Set of is reported in connection with this paper. Rationalizable Strategies (SRS) and their associated payoff vectors are sought for THU-4-F each of the players. This is done using Pareto-based best replies of the opponent, Contributed Session: MCDM for with an iterative search technique that Strategic, Tactical and Operational accounts for mutual-rationalizability. The Decisions proposed mutual-rationalizability approach Thursday 14:20-16:00 - Room: Basilica results with trade-off information at the end Cistern of the first stage. In the second stage, Multi- Criteria Decision Analysis (MCDA) techniques Chair: Alexander Engau are used to select a strategy out of the obtained SRSs based on the trade-off 1. World Mega Trends and Their information that is revealed in the first stage. Implications for MCDM Research To demonstrate the proposed solution Hannele Wallenius, Aalto University, concept a MOG version of the Traveling Finland, [email protected] Salesperson Problem (TSP) is employed. The latter game type (TSP MOG) amalgamates *Jyrki Wallenius, Aalto University, two known versions of the TSP including the Finland, [email protected] competing TSP and the selective TSP. The We overview several of the technology mega demonstrations include several case studies trends and other trends, which are of interest with a relative small network of cities, which from the MCDM perspective. We discuss, allows full sorting that results in the exact what role MCDM could play in this SRSs. The main conclusion from this study revolutionary development. We also discuss is that the proposed solution concept, as how the mega trends change the field of amalgamated with the suggested MCDA MCDM. Digital technology is making rapid techniques, provides players of MOGs with a advances. The implications for people, novel approach to support justified strategy companies, and societies are pervasive. The selection based on trade-off information. envisioned changes will bring about: (1) digital connectivity, independent of time and place, and (2) tools for quickly analyzing vast amounts of digital data. In the World Economic Forum’s recent report, the 154 changes are grouped into six “mega-trends”. the learning robots have ‘emotions’ and We will focus on the following three. 1. The empathy? New opportunities for our field are Internet – world’s access to the Internet will offered by platform economies, where supply continue improving; people’s interaction with and demand meet. The preferences of those it will become more ubiquitous 2. Big data who demand services and who offer services and Artificial Intelligence – the ability to must be matched. We can envision several access and analyze huge amounts of data 3. novel application areas for MCDM, such as The sharing (or platform) economy These online auctions, voting advice applications, trends will greatly impact our lives, ride-sharing apps, accommodation-sharing businesses, and governments - even sites. The environment is one of the most universities - all around the world. As the popular application areas for MCDM tools. World Economic Forum’s Report astutely These problems require DMs to consider observes, our lives are increasingly being multiple criteria and complex tradeoffs. An driven and enabled by software. The important area for MCDM is helping design potential of the digital technology is huge, environmental and energy policy. Moreover, both in enhancing traditional industrial a novel application area is investment processes (robotics), and even more decisions to be based not only on expected importantly in generating novel digital returns and volatility, but a third dimension: services. Many aspects of healthcare are sustainability, ethics. We conclude our also benefiting tremendously from new presentation by describing our views on how technologies. The digital revolution has we see MCDM changing as a result of these begun. Besides technology mega trends, trends. there are other highly important mega trends. These mega trends, unlike 2. Using FICO Xpress for Solving technology mega trends, are generally Large-Scale Multi Objective perceived as challenges or threats to Combinatorial Optimization humankind. Some of them are discussed in Problems in Finance PwCForesight#megatrends and by the World Economic Forum, for example: climate *Sebastien Lannez, FICO Xpress change, concern for environment. We Optimization, United Kingdom, discuss the above mentioned technology [email protected] mega trends and the non-technology mega trend related to climate change from the With the FICO Decision Optimizer software, above World Economic Forum’s list. What powered by FICO Xpress solvers, a strategy role can MCDM play in them? How can analyst can define and optimize complex MCDM help? What MCDM concepts will be decision problems using an intuitive useful? The ubiquitous nature of internet graphical workflow. In this talk, we will show will present new opportunities for MCDM how an analyst user without prior experience scholars. We need new kind of decision of optimization can create a loan amount support for billions of users who shop online. pricing optimization application and solve The support in an online environment must, complex multi objective optimization however, be simpler than traditional MCDM problems. The presented example business tools. Search engines used to find desired problem aims at determining the best loan products or services, but (price-only) search amount and price to be offered to the engines are not necessarily good enough in customers in a lender’s portfolio. In order to differentiating among offers. There is a need comply with regulatory constraints and for multi-attribute search engines. Moreover, company policy the solution must satisfy popular recommender systems should better certain budget and loss limits. These capture people's preferences. Big Data calculations are defined by various artefacts, needs optimization tools to help DMs, data like PMML probability models or equation alone is not sufficient. Do we want to components, and then combined and linked delegate important decisions to robots? The together to form the Decision Impact Model, MCDM spirit has been to highlight the a graphical model representing the full importance of the human decision maker. decision making process that leads to the We feel that it is important that technology is offering, which can also be considered as a seen as augmenting human capabilities, not Bayesian Networks of causal probabilities replacing them. How do we make sure that which calculate the expected decision

155 outcome values. This Decision Impact Model [1] Haimes, YY, Ladson, LS & Wismer DA, is automatically transformed into "Bicriterion formulation of problems of optimization problems, simulation routines integrated system identification and system or scoring algorithms, without any optimization." IEEE Transactions on Systems intervention from the end user. Solutions Man and Cybernetics, vol. 3, pp. 296, 1971 generated by the software can then be [2] Kirlik G, Sayin S, "A New Algorithm for compared using the built-in dashboards Generating All Non-dominated Solutions for making it simple to benchmark business-as- Multiobjective Discrete Optimization usual or challenger solutions from a Problems." European Journal of Operational simulated decision process against Research, Vol. 232, pp. 479—488, 2014 optimized solutions generated by the solver. The optimization problem that is 3. On Combining Explainable automatically produced by Decision Artificial Intelligence and Interactive Optimizer is a Generalized Assignment Multiobjective Optimization in Data- Problem. Each account in the portfolio is Driven Decision Support assigned a treatment which corresponds to a set of predefined actions. The global *Jussi Hakanen, University of Jyvaskyla, constraints of the GAP are used to model Finland, [email protected] resource limits like budget or available Vesa Ojalehto, University of Jyvaskyla, offers. A typical instance of this application Finland, [email protected] has 150 treatments, 1M accounts and Mirka Saarela, University of Jyvaskyla, between 5 to 20 global constraints. The two Finland, [email protected] functionalities we will present during this talk Sami Äyrämö, University of Jyväskylä, allow the user to explore the objective function value space and discover the Finland, [email protected] relationships between various objectives by Nowadays, many decision making processes solving multiple objective combinatorial are driven by data and, thus, the term data- optimization problems: • In the first driven optimization has been widely used. approach, the user specifies what are the Data can be coming, e.g., from different objectives or constraints to explore, defines sensors due to rapid rise of Internet of ranges and the exploration step size and lets Things, experimental measurements or the optimizer search for optimal solutions in social networks. The amount of available the partitioned feasible space. This approach data is often huge and it poses challenges for can be seen as a simplified version of the decision making in i) how to find and apply epsilon constraint method described by relevant data for the problem at hand and ii) Haimes, Ladson and Wismer [1] in which the how to use that in supporting decision epsilons are uniformly sampled over a making. On top of that, the problems often distribution defined by the end user. The have multiple conflicting criteria that need to optimization problem is solved for every be optimized simultaneously, thus, requiring partition and the optimal solutions are multiple criteria decision making (MCDM) displayed on a two-dimensional efficient techniques in finding a most preferred frontier graph for which the two axes are solution. Machine learning (ML) tools are taken from the set of objectives. We show often essential in building the optimization that this predefined partitioning of the problem from the data. In addition, ML can feasible space is a tractable approach for be used to learn decision maker's (DM’s) large optimization problems. • With the preferences in order to propose promising second method the user can choose to apply solution candidates during the solution a more dynamic approach based on an process. If the DM does not understand why implicit enumeration of all possible epsilon certain solutions are proposed, it may hinder values as proposed by Kirlik and Sayin [2]. the DM in trusting those recommendations This latter approach ensures that all non- resulting in not considering them at all. dominated solutions will be found but is Therefore, the ML tools used should also give computationally more demanding. We will an explanation why the solutions are give preliminary results comparing the time proposed to the DM . These methods belong and memory requirements of the two to explainable artificial intelligence which is approaches. an emerging research field due to high popularity of applications of artificial

156 intelligence. So far, explainable ML methods 4. Multicriteria Project Prioritization have been used to explain performance of in Transportation Asset Management ML algorithms but they have not been used much as a decision support for MCDM (i.e. in *Alexander Engau, Dalhousie a prescriptive analytics context). By University, Canada, explaining their reasoning to the DM, the ML- [email protected] based decision support tools become more easily trusted and accepted by them. This Transportation systems all over the world means that when the DMs understand better continue to grow rapidly and to become the reasons behind the decision support, increasingly complex. To better program and they are equipped to make more transparent respond to the many resulting challenges, and trustworthy decisions. This is especially Transportation Asset Management (TAM) the case when dealing with multiple deals with the planning, building, operating, conflicting objectives, where understanding maintaining, upgrading or expanding of the the trade-offs between the objectives is underlying transportation infrastructure and crucial. In this paper, we discuss the its physical assets including roads, bridges challenges of combining explainable ML with and any other transportation facilities. interactive multiobjective optimization in a Hence, in their most general form, TAM goals data-driven context. By using an example are to optimize overall system performance case study, we show how these two distinct including cost effectiveness and efficiency, approaches can be combined and what kind resource allocation and utilization as well as of issues must be considered in order the the general satisfaction of all users and combination to be effective. To our system stakeholders. It follows that TAM is knowledge, this has not been done before. inherently multi-criterion in nature so that its Typical to interactive methods is that the decisions and any related decision-making human DM actively participates in the procedures should ideally follow best solution process and provides preference business and engineering practices and be information when the most preferred conducted based on quality, relevant and solution is searched. The challenges include credible information with well-defined 1) which way to utilize ML within the objectives for a meaningful tradeoff and optimization process, 2) what kind of decision analysis. Following a general explanations to provide for the DM, and 3) discussion of transportation asset how to present the explanations to the DM . management in practice, this presentation The first challenge deals with identifying the then will focus specifically on one “real- role of ML as a part of the whole solution world” situation based on a recent process and what kind of ML tools to use. It collaboration with a major transportation is commonly known that there exists a trade- agency in the United States. In agreement off between the performance and with the mission, vision and general goals of explainability of ML models. For example, its recent strategic management plan, we will deep learning with neural networks has begin to briefly outline the underlying recently become popular ML approach for objectives hierarchy which includes system complex data. While those models have a performance and general organizational good performance, their explainability is very excellence, safety and health, stewardship low. On the other hand, decision trees have and efficiency as well as sustainability, high explainability but may not perform so livability and economy. In particular, having well with complex data. Secondly, the type of been invited to review and further propose explanations depends on the ML models associated multi-objective decision analysis used as well as the application considered (MODA) approaches for its optimal resource and they can be, e.g., visual or descriptive. allocation and project prioritization, we will Finally, how to communicate the highlight a few specific lessons we have explanations to the DM is also important and learned. First, we will shortly revisit the requires a graphical user interface (that is original suggestion to merely use a standard also an important element in interactive cumulative benefit for a classical benefit- multiobjective optimization methods) cost knapsack heuristic and comment on its utilizing, e.g., techniques from visual well-known drawbacks and perceived analytics to communicate the message. disadvantages in comparison to some of its other more positive benefits. Second, we will

157 continue in more detail to also describe two closely related, alternative but actual MODA approaches that are based on multi- objective goal programming and data envelopment analysis, respectively. Third, we will compute and compare their final compositions of several optimal project portfolios using an experimental but realistic data set of 40 candidate projects with a total cost of approximately $855.9 million dollars under different budgets. Finally, and in conclusion, we would also be delighted to entertain additional comments or questions for the further discussion of each of these current MODA approaches and their possible use, value and ongoing improvement.

158

model preference parameters. This can be performed indirectly by inferring the FRIDAY parameters from stated preferences, as done in previous research papers. Learning an RMP model from stated preferences FRI-1-M proves however to be computationally Society Award Talks extremely costly and can hardly be put in practice using state of the art algorithms. We Friday 9:30-12:00 - Room: Galata Tower propose a Boolean satisfiability (SAT) formulation of the inference of an RMP Chair: Jyrki Wallenius model from a set of pairwise comparisons, FRI-2-A which is much faster than the existing algorithms based on mathematical Invited Session: New Approaches in programming. We present the SAT clauses MCDM that aim at inferring the model and explain them. The SAT formulation is tested on Friday 13:00-14:40 - Room: Galata Tower artificial data sets. We set up an Chair: Gülşah Karakaya experimental framework and assess the computing time and the performance of the 1. Ranking with Multiple Reference formulation in generalization. In the Points : Efficient Elicitation and experiments, we vary the size of the learning Learning Procedure set and the size of the model that has to be learn. Finally, we present a MAX-SAT Khaled Belahcene, CentraleSupélec, formulation which allows inferring a RMP France, model when the learning set is not fully [email protected] compatible with such a model. The MAX-SAT Vincent Mousseau, CentraleSupélec, formulation is tested on artificial datasets. France, 2. PROMETHEE Approaches for [email protected] Wassila Ouerdane, CentraleSupélec, Medical Decision Support Under France, Uncertainty [email protected] *Melodi Cebesoy, Hacettepe University, Marc Pirlot, University of Mons, Turkey, Belgium, [email protected] [email protected] *Olivier Sobrie, University of Mons, Ceren Tuncer Şakar, Hacettepe Belgium, [email protected] University, Turkey, In the context of multicriteria ranking [email protected] problems, we consider a ranking procedure Barbaros Yet, Hacettepe University, based on reference points recently proposed Turkey, [email protected] in the literature. This method, known as Musculoskeletal diseases have a high Ranking with Multiple reference Points burden on healthcare resources in Turkey (RMP) makes use of the following preference due to increasing age of the population. parameters to specify the decision maker Therefore, we aim to develop a decision judgment: (i) a set of reference points and (ii) support tool that can support and validate an importance relation on criteria coalitions. the decision makers’ (doctors or Using the RMP method, alternatives are physiotherapists) choices and offer other compared to reference points considered alternatives to patients who are far from the successively in a certain order. An alternative physiotherapy centres. Musculoskeletal is considered better than another if the diseases have a variety of treatment options coalition of criteria on which it is better than such as medical interventions and physio a reference point is more important than the therapy applications like surgery, manual coalition of criteria on which the other is therapy and orthoses. Several factors are better than the same reference point. considered when assessing the outcomes of Implementing the RMP method in a real these treatments such as pain, side effects, world decision problem requires to elicit the cost and convenience for the patient. We

159 handle these factors as criteria and the In each iteration, we find the assignment treatment options as our potential frequency of alternatives for each category alternatives. Accordingly, in order to detect using the values of the binary variables the most effective treatments for patients, gathered from the MIPs and calculate the we use a Multi Criteria Decision Making probability of an alternative to be placed in a (MCDM) approach. In our approach, category based on these frequencies. We treatment options can have probabilistic then use the information theoretic measure, outcomes in some criteria. To represent this entropy, in the selection of the alternative uncertainty, Bayesian Network (BN) models that will be placed to a category by the are used. A BN is a graphical probabilistic decision maker. The entropy concept fits well model that represents relations of variables to our measurement of uncertainty about the and their probability distributions. In our categories of the alternatives since the aim problem, output of the BN shows the is to ask the decision maker an alternative probabilistic outcomes of decision that will yield much information to the alternatives in the criteria considered for a decision process. We test the performance of given patient. After obtaining the outcomes our approach on an example problem from from BN, we use the well-known outranking the literature and the results show that it method PROMETHEE. To obtain the weights works well. of criteria used in PROMETHEE, we consider carrying out Analytic Hierarchy Process with 4. Comparison of Solutions under a experts in the field of physiotherapy. To be Weighted Tchebycheff Function able to work with probabilistic outcomes, we *Gülşah Karakaya, Middle East first make use of Monte Carlo Simulation to generate a large number of discrete Technical University, Turkey, scenarios. Using these scenarios, we work [email protected] with both PROMETHEE I and II to evaluate the Murat Köksalan, Middle East Technical treatment options. PROMETHEE I produces a University, Turkey, partial ranking of alternatives, whereas [email protected] PROMETHEE II produces a complete ranking. We apply our approach to evaluate nine In this study we address the comparison of a treatment options with six criteria. Working number of solutions that are defined by with probabilistic outcomes of criteria, our multiple objectives and evaluated with a approach is able to provide decision support weighted Tchebycheff function. Tchebycheff that can handle this uncertainty. functions can be used for a variety of purposes when multiple criteria are 3. An Interactive Sorting Approach considered. One such use is the evaluation Based on Information Theoretic of an approximation of a Pareto set. It is not straightforward to find the true Pareto set Measure especially for multi-objective combinatorial *Ali Özarslan, Middle East Technical optimization problems. Heuristics have been University, Turkey, [email protected] developed to approximate the Pareto set. To Gülşah Karakaya, Middle East Technical measure the quality of approximate Pareto University, Turkey, [email protected] sets and to compare such sets with each other, there are some performance In this study, we develop an interactive indicators such as hypervolume measure, approach for sorting alternatives. We epsilon indicator, and integrated preference assume that the preferences of the decision function (IPF) measure. A Tchebycheff maker are consistent with an additive function-based IPF measure can be used to function. We assign worst and best possible estimate how well a set of solutions categories for each alternative and narrow represent the Pareto set. It measures the down these category ranges using mixed quality of the representation well but is integer programming (MIP) iteratively. We computationally demanding especially as the utilize binary variables to assign the problem size gets larger. We improve the alternatives for which the classes are not computation of the IPF measure to make it known exactly. We incorporate the worst and practical for complex problems. We develop best possible category information to the the necessary theory to define some MIPs whenever new information is obtained. relations of the solutions in the weight space

160 so that the IPF calculation can be done the uncertainty on IT project performance efficiently. The theory not only facilitates the management have not been addressed in practicality of the IPF calculation but also detail. This brings about many problems enables to incorporate the preferences of a especially unmet cost, time, and quality decision maker whose preferences can be goals. Thus, most of the projects are failed approximately represented by a Tchebycheff and IT companies waste their resources. function to converge preferred regions of the Therefore, researchers and IT practitioners solution space in an efficient manner. We try to find new ways of measuring the demonstrate our approach on example performance of IT projects with the help of problems with two, three, and four effective criteria and improve the results by objectives. focusing on the root causes of performance issues. Furthermore, since most IT FRI-2-B companies place emphasis on meeting cost, Contributed Session: Fuzzy sets and quality, and time goals, there is a need to approaches enhance the set of performance criteria in such a way that it covers other fundamental Friday 13:00-14:40 - Room: Maiden's Tower performance factors (i.e., ROI, net profit margin, customer and team satisfaction, Chair: Sait Gül developing skills etc.) especially in uncertain 1. IT Project Performance environment. This paper presents a novel Management in Fuzzy Environment: methodology to measure the performance of IT projects in fuzzy environment by A Case Study combining Balanced Scorecard (BSC) used *Ayfer Başar, Istanbul Technical for performance management and Hesitant University, Turkey, Fuzzy Sets (HFS) used for subjective [email protected] weighting of IT project performance factors. BSC is an internationally accepted method to The project is described as a temporary model the performance management of IT discrete endeavor to create a unique projects by its balanced and product, service, or result. Project multidimensional structure. On the other management involves the processes about hand, HFS is a new area in the fuzzy planning, executing, controlling and closing literature. In this study, firstly, we to deliver the project outcomes (PMI, 2017). hierarchically model the IT project Assessing the performance of projects performance problem with the help of BSC accurately becomes challenging in all the structure, and then find the importance industries and the failure directly affects the values of 4 perspectives (finance, customer, sustainability of the companies. Therefore, internal business processes, and learning- organizations look for the efficient growth) and 12 critical success factors techniques to measure and increase the (return on investment, net profit margin, and project performance. Traditionally, project meeting the budget for finance; customer performance was calculated based on the satisfaction, contribution to the market fulfillment of the cost, time, and quality share, and creating the new market for the goals. Later, this approach was found customer; meeting customer needs, meeting insufficient and a set of new criteria are schedules, team efficiency for the internal proposed for the project performance processes; team satisfaction, developing calculation (i.e. meeting the requirements of skills, and contributing to the retention of quality, safety, customers satisfaction, team members for learning and growth) with changes in the scope, innovation etc.) In the help of expert judgment and HFSs. We addition to a wide range of criteria, different experimentally show that finance is the most techniques have been applied to measure important perspective while customer is also the success of projects in academic and fundamental in terms of achieving IT vision professional studies (i.e. Earned Value and mission. Furthermore net profit margin, Management - EVM, Earned Schedule - ES, creating the new market, meeting schedules, and Earned Duration Management – EDM). and team satisfaction are the most Although the literature on project important critical success factors in terms of performance management is rich; the finance, customer, internal business frameworks and solution methods handling processes, and learning-growth respectively.

161

The results of the case study demonstrate be provided and there will be savings in the that the performance of 7 IT projects are transportation and storage and labor costs. higher than 60%, and there is only one With these features, VRP applications have project with its performance below than 30% been taking more and more place in the among 68 IT projects completed in a Turkish health sector. When the domestic and company in 2018. The outputs of this study foreign literature on VRP studies in the field are found acceptable by senior IT managers, of health are examined; ambulance, dialysis, project managers, customers, and team blood transfusion and home care services members of the company. have been observed to be VRP studies. In this respect, the aim of this study is to 2. Selection of the Appropriate determine the best VRP’s solution method Solution Method for the Home which can increase the efficiency of HHC by Health Care Vehicle Routing Problem using Fuzzy Analytic Network Process (FANP). by Fuzzy Analytical Network Process: This model includes both criteria (number of A Case Study of Erzincan patients, cost of solution, solution time, the number of vehicles and employees) and the İskender Peker, Gümüşhane University, alternatives (definite, intuitive and meta- Turkey, heuristic solution methods). The analysis [email protected] was carried out on the vasis of information Gökhan Çaybaşı, Erzincan Binali Yıldırım shared by Erzincan Provincial Directorate of Health and actual data sets obtained within University, Turkey, the scope of Personal Data Protection Law. [email protected] According to results, the most important Gulcin Buyukozkan, Galatasaray criterion is solution time. Meta-heuristic University, Turkey, solution methods is the appropriate method [email protected] for the home healthcare vehicle routing problem. In recent years, logistic activities attract more attention with their increasing 3. Usage of Entropy-Based Objective importance in every sector. Health is one of Weighting in Neutrosophic Multiple the most important of these sectors. According to the World Health Organization Attribute Decision Making (WHO), “Health logistics is an important part *Sait Gül, Beykent University, Turkey, of technical assistance in emergency [email protected] situations; it is to manage the transport of infectious substances for laboratory tests, Multiple attribute decision making (MADM) including the maintenance and distribution approaches focus at selecting the best of stocks, and to provide coordination during alternative from a set of alternatives by the outbreaks”. The health institution brings considering different attributes. Since together the client or the patient and the evaluation of alternatives is based on the health care team or the doctor and takes on importance of the attribute which are a role in preparing the optimal environment represented by weights, decision analysts for the treatment to be implemented. The should choose a proper weighting method. way in which health services are presented is There are two basic types of weighting. The one of the most important indicators of socio- first one is subjective weighting (for instance, economic development levels. Home Health pairwise comparison of AHP) that consults Care (HHC) in Turkey aims to give health the expertise of decision makers (DMs) with services to people who are bedridden in their the aim of revealing the hidden information residence. The health sector is a sector that in their conscious and showing them with can cause irreparable results if the services numbers. The second weighting is called delay. For this reason, the logistic objective weighting. This type ignores the infrastructure should be strong and should DMs’ attribute prioritization and just looks be used effectively in health services. With the alternative points which are measured the application of Vehicle Routing Problem with respect to attributes. In literature, some (VRP) in many areas of the health sector, in drawbacks are specified for subjective which time is the biggest limitation, weighting such as risks that may be improvements in the sector of the route will originated from self-seeking DMs or possible long time periods that are required for data 162 collection from DMs. To overcome them, two different hypothetical decision making objective weighting methods are developed. problems in order to present the proposed In real life, decision problems and method’s usability and efficiency. information about alternatives can be Exponential entropy by Ye and Cui (2018), 4 incomplete, indeterminate and inconsistent. different methods developed by Ye and Du To deal with this problem, there are some (2017), a method from each of Aydoğdu propositions in literature. Intuitionistic fuzzy (2014), and Majumder and Samanta (2014) set (IFS) which was introduced by Atanassov are used for the calculation of objective (1986) as an extension of Zadeh’s fuzzy set weights. The results of applications show theory (1965) was generalized to a new that the entropy-based objective weighting is concept called neutrosophic set (NS) theory properly useful in IVNS based N-MADM. An by Smarandache (1998). The membership ANOVA (analysis of variance) test is utilized degree is proposed as a real number in unit for the comparison of the weight sets interval by Zadeh and two degrees are generated by 7 different entropy-based proposed to represent uncertain information methods. in IFSs: degree of membership and degree of non-membership. NS is described 4. Objective Weighting of Decision- independently by truth, falsity, and Makers and Attributes in indeterminacy membership degrees: A={x, Neutrosophic Group Decision TA(x), IA(x), FA(x) | x ϵ X}. In literature, the Making: A Review single valued NS (SVNS) and interval valued NS (IVNS) are proposed as the instances of *Sait Gül, Beykent University, Turkey, NS. TA(x), IA(x), FA(x) ϵ [0,1] are the [email protected] membership degrees of a SVNS, Each MADM method requires information consecutively. In IVNS, the membership about weights representing the priorities of degrees are represented as intervals. In attributes. Similarly, in group decision existing literature, there are many papers making (GDM) decision makers (DMs) can studying the neutrosophication of classical usually possess different levels of expertise MADM methods (N-MADM). For example, which can be measured by weights of them. Basset et al. (2018) proposed a In real-time problems, both types of weights neutrosophic AHP; Kour and Basu (2017) can be incompletely known or completely developed a neutrosophic TOPSIS and Liang unknown. There are two basic weighting et al. (2017) constructed a SVTN-DEMATEL procedures: subjective and objective module. To our best of knowledge, there is no techniques. In real life applications of MADM objective weighting methodology proposition and GDM, information required can be in literature and all the current articles apply incomplete, indeterminate and inconsistent. the subjective weighting of attributes. To handle this issue, the neutrosophic set Entropy is an important mathematical tool for (NS) theory was developed as an extension measuring uncertain information. Entropy- of fuzzy sets. NS is described independently based objective weighting is based on an by truth, falsity, and indeterminacy idea: an attribute is more important if there membership degrees: A={x, TA(x), IA(x), FA(x) is a greater dispersion in evaluations of | x ϵ X}. In this study, the aim is to make a alternatives for this attribute. According to brief overview of objective weighting of this definition, the dispersion of the data in attributes and DMs in N-GDM. After an the same attribute can be a measurement of extensive research and filtering process on its importance. There are many entropy scientific databases, such as Google Scholar measurement approaches in NS literature and Web of Science, 27 papers handling and various examples are given below. If the objective weighting issues were found. By a entropy of an attribute j is represented by Enj, detailed analysis, brief inferences which are its weight can be calculated by the formula of given below are made. In literature, 26 wj = (1-Enj) / (Σj(n-Enj)) where n is the papers of 27 utilized any objective attribute number of alternatives. In this study, we weighting methods and eight different propose the usage of entropy measurement approaches were noticed. The most utilized in determining the attribute weights for the one is Maximizing Deviation by 11 papers first time in literature with a focus on (42%). Entropy takes the second place with 5 objective weighting in N-MADM. 7 entropy papers (20%). In the 3rd place, there are methods for IVNSs are considered for solving

163 optimization models from 4 papers (15%). FRI-2-C The remaining methods are Grey system, mean-squared deviation, variation Contributed Session: Advances in MCDA coefficient, TOPSIS, and multi-objective Friday 13:00-14:40 - Room: Dolmabahçe optimization. Deriving DMs’ weights Palace constitutes a relatively new domain in GDM area. In N-GDM literature, only 4 papers used Chair: Ana Sara Costa an approach regarding weighting of DMs. TOPSIS-based approaches were used by 3 1. Decision Making Procedure Using papers and one of them used variation Strict Preference Information for coefficient method simultaneously. One Group Decision Based on Multi- paper utilized a score-function based Attribute Utility Analysis approach. While 5 papers accepted the DMs’ weights as given, 4 papers appealed Tomohiro Hayashida, Hiroshima qualitative group consensus. Also, we found University, Japan, 14 MADM applications in N-GDM literature. [email protected] The mostly used ones are TOPSIS by 9 (33%), *Ichiro Nishizaki, Hiroshima University, GRA by 7 (26%) and VIKOR by 3 papers Japan, [email protected] (11%). Among 4 papers which considered the Shinya Sekizaki, Hiroshima University, weights of DMs, 2 papers selected VIKOR, 1 Japan, [email protected] paper chose an integration of TOPSIS and Masato Ono, Hiroshima University, QUALIFLEX, and the last one used a score function based method. As additional Japan, [email protected] findings, the N-GDM studies used various Multi-attribute utility analysis (MAUT) types of information as instances of NS. 11 constructed by Keeney and Raiffa (1976) is papers (41%) worked with single-valued NS, a method for decision making for multi- and 6 papers (22%) utilized interval-valued criteria decision-making problem, which NS. Each paper performed an application of models the trade-off relationship between their proposed methodologies but only 4 of attributes considering the preference them worked on real problems (15%). Finally, structure of a decision maker. The multi- only 10 papers (37%) are published by attribute utility function consists of a utility SCI/SCI-E indexed journals. While the function for a single attribute and a scaling number of papers which appeared in constant representing a weight for a plurality SCI/SCI-E journals is just 2 before 2017, of single attributes. The scaling constant remaining 8 papers have been published quantitatively expresses the trade-off during the last 2 years. Consequently, it is relationship between decision-making seen that the N-GDM literature requires more person's attributes, and in the procedure objective methodologies since the number of proposed by Keeney and Raiffa, it is uniquely papers studied these weighting issues is few identified by indifferent experiments. Based and the applications of developed on the identified utility function, the decision methodologies are limited. 23 papers (85%) maker can rationally choose one of the most introduce only illustrative applications, not preferred alternatives. However, when there real ones. It is obvious that it is required to are many attributes to estimate or when show their applicability in real life problems. collective decisions are made by multiple The number of papers published in SCI/SCI- stakeholders it is difficult to answer E indexed journals has been increased after consistently through multiple questions to 2017. So, this finding shows the increasing identify the preference structure. Since all reliability of this novel decision making decision makers do not have expertise, utility domain. All these findings can provide functions cannot be uniquely identified. guidance to the future researches on Hayashida et al. (Hayashida et al., 2010, objective decision making in neutrosophic 2011) show the usefulness of MAUT in environment. practical problems. As an extension procedure of MAUT Nishizaki et al. (2013) have proposed a decision procedure such that the decision maker or the corresponding stakeholder answer the strict preference relationship between some pairs of virtual

164 alternatives which can be easily answered to It is important in particular the field of estimate the scaling constants of the transport project appraisal, as many projects corresponding multi-attribute utility function. fail to be implemented because of lack of Furthermore, in case that scope of the support from one or more key stakeholders. scaling constants cannot be determined In contrast to many multi-criteria decision based on a single exact relationship between analysis (MCDA) methods, multiple the pairs of alternatives, a new questionnaire stakeholders can use different criteria trees is generated based on the result of the in MAMCA (with their own preferences). The answers. As a procedure based on such idea, steps of the classic MCDA process include: a neural network MAUT (NN - MAUT) definition of problems, alternatives and (Nishizaki et al., 2014) is proposed which a criteria, analysis of alternatives, neural network is applied to estimate the determination and analysis of scores, and scaling constants. Here, the weights and the the drawing of conclusions. Unlike classic thresholds of a neural network are updated MCDA, MAMCA involves stakeholders after by genetic algorithms. MAUT Excel defining the alternatives. The stakeholder (Hayashida et al., 2013) and MIDASS (Seo et analysis is conducted, stakeholders are al., 2004, 2007) have been constructed as identified and the criteria tree for each of decision support systems based on MAUT. them can be constructed. In this way, This study proposes a procedure to select an MAMCA explicitly takes into account their alternative rationally even if a decision own objectives. In the end, the advantages maker does not have enough expert and disadvantages of each of the proposed knowledge. Nishizaki et al. (2013) assumed scenarios are presented and possible that a decision maker is asked some strict consensus scenarios are discussed. Yet in preference relationships between some pair some cases, it is challenging to reach a of the virtual alternatives, and if there is no consensus based on MAMCA results. In this contradiction in the answers, the scaling contribution, we propose a way to help constants are uniquely determined. This facilitators and stakeholders to find a study proposes a procedure to select consensus utilizing the PROMETHEE method alternatives by estimating their distribution if in MAMCA. Doan and De Smet recently the answer includes some contradiction. developed an alternative weight sensitivity Note that the procedure of Nishizaki et al. analysis for PROMETHEE based on mixed (2013) deals with multi-attributes utility integer linear programming (MILP). MILP functions which are defined in multiplicative avoids altering all the weights and reduces form, however, this study deals with utility the distance from the original weights functions which are defined in additive form compared to classic weight stability intervals to simplify the calculation procedure. (WSI) (which can find the minimum weight modification to change other alternatives to 2. A Method of Reaching Consensus be ranked first). It can be applied in the with the Multi-Actor Multi-Criteria MAMCA methodology to offer a consensus Analysis (MAMCA) methodology between different stakeholders by taking the inverse optimization point of view. The MILP- *He Huang, Vrije Universiteit Brussel, MAMCA method allows computing two Belgium, [email protected] possible indicators R1 and R2 (with two Cathy Macharis, Vrije Universiteit different ways to deal with the maximum Brussel, Belgium, weight modification), which indicate the [email protected] distance among all stakeholders for each Yves De Smet, Université libre de alternative to reach the consensus. We apply Bruxelles, Belgium, this new approach on two MAMCA project [email protected] cases. Then we examine how to integrate this model into the existing MAMCA methodology Anh Vu Doan Nguyen, Université libre and visualize the results with the MAMCA de Bruxelles, Belgium, software. [email protected]

The multi-actor multi-criteria analysis (MAMCA) is a methodology which allows for the inclusion of multiple stakeholders in the process of evaluation and decision making. 165

3. Stochastic Analysis Applied to a at hand. The application of the proposed Multiple Criteria Hierarchical comprehensive method is illustrated through a didactic example. Acknowledgments: This Nominal Classification Method work was supported by national funds based on Similarity and Dissimilarity through Fundação para a Ciência e a *Ana Sara Costa, Universidade de Tecnologia (FCT) with reference Lisboa, Portugal, UID/CEC/50021/2019. Ana Sara Costa acknowledges financial support from [email protected] Universidade de Lisboa, Instituto Superior Salvatore Corrente, University of Técnico, and CEG-IST (PhD Scholarship). Catania, Italy, [email protected] 4. The Method of Multi-Attribute Salvatore Greco, University of Catania, Non-Inferiority and Superiority Italy, [email protected] Scores Heavy Averages (MANISSHA) José Rui Figueira, Technical University *Nourchen Moumni, University of Sfax, of Lisbon, Portugal, Tunisia, [email protected] [email protected] José Borbinha, Universidade de Lisboa, Abdelwaheb Rebaï, University of Sfax, Portugal, [email protected] Tunisia, [email protected] In multiple criteria nominal classification In the current work, we show how to rank pre- problems, the assignment of an action may specified alternatives by employing what will include a set of possible categories, instead be referred to as the method of Multi- of only one, raising the challenge of how to Attribute Non-Inferiority and Superiority support the Decision Maker (DM) in Scores Heavy Averages (MANISSHA). This analyzing the accuracy of the assignment of new method to multi-attribute ranking is each action. In this work, we propose a founded on the three pillars: (1) non- robust nominal classification method using inferiority and superiority scores, (2) heavy criteria structured in a hierarchical way, aggregation operators, and (3) Einstein sum adopting an integrated approach. Stochastic operation. More precisely 1. The aforesaid Multicriteria Acceptability Analysis (SMAA) non-inferiority and superiority scores of and the nominal classification method CAT- competing alternatives are produced using SD (CATegorization by Similarity (crisp or valued) outranking relations. 2. Dissimilarity) are put together to deal with The aforementioned scores are aggregated nominal classification problems with using the heavy weighted average (HWA) hierarchical criteria and imprecise weight and/or the heavy ordered weighted average information. This proposal follows four main (HOWA) operators. 3. The values returned steps: (i) application of Multiple Criteria by the above operators are normalized in a Hierarchy Process (MCHP) to CAT-SD; (ii) new and unfamiliar manner and the resulting application of the imprecise SRF method, for values are merged using the Einstein sum each category, while the hierarchy of criteria operation. 4. The competing alternatives and the possible interactions effects in are ranked from most to least preferred criteria pairs are considered; (iii) according to their Einstein sum operation combination of SMAA and the hierarchical output values. The proposed method is CAT-SD (SMAA-hCAT-SD) by sampling sets of illustrated using a multi-attribute ranking parameters compatible with the preferences problem and the obtained ranking results are of the DM previously provided; and (iv) obtain compared to those yielded by the existing a final classification of the actions that fulfills superiority and inferiority indexes based a set of requirements given by the DM, by ranking methods: SIR-SAW, SIR-TOPSIS, SIR- adoption of a novel procedure that allows VIKOR, and SISINA. Keywords: Multi- considering the probabilistic assignments attribute ranking, Non-inferiority score, provided by SMAA-hCAT-SD. The results Superiority score, Einstein sum operation, consider robustness concerns, while take Outranking relation. into account the whole set of criteria or a certain node in the hierarchical structure. Thus, the DM can have a better and deeper understanding of the classification problem

166

FRI-2-E their new projection directions point towards areas of the PF which are not so well Contributed Session: Evolutionary approximated. In order to show the Algorithms and Multiobjective effectiveness of A-GWASF-GA, we compare it Optimization with NSGA-III [2, 3], MOEA/D [4], MOEA/D- AWA [5] and the original GWASF-GA, Friday 13:00-14:40 - Room: Hagia Sophia considering ten problems with three, five, Chair: Michael S. Bittermann eight and ten objectives and with twenty and fifty decision variables. To evaluate their 1. A-GWASF-GA: The New Version of performance, we use the IGD metric [6]. The GWASF-GA to Solve Many Objective results of the computational experiment Problems demonstrate the good performance of A- GWASF-GA in the novel many-objective *Sandra Gonzalez Gallardo, University optimization benchmark problems of Malaga, Spain, [email protected] considered. Mariano Luque, University of Málaga, References 1. R. Saborido, A.B. Ruiz, M. Spain, [email protected] Luque, Global WASF-GA: An evolutionary Rubén Saborido, University of Málaga, algorithm in multiobjective optimization to Spain, [email protected] approximate the whole Pareto optimal front, Ana B. Ruiz, University of Málaga, Evolutionary Computation 25 (2017) 309– Spain, [email protected] 349. 2. K. Deb, H. Jain, An evolutionary many-objective optimization algorithm using A new version of the evolutionary algorithm reference-point-based nondominated sorting based on GWASF-GA [1] is proposed in this approach, Part I: Solving problems with box work. GWASF-GA is an aggregation-based constraints, IEEE Transactions on algorithm which uses the Tchebychev metric Evolutionary Computation 18 (2014) 577– plus an augmentation term as fitness 601. 3. H. Jain, K. Deb, An evolutionary function and two reference points (the many-objective optimization algorithm using utopian and nadir points) to classify the reference-point based nondominated sorting individuals according to a set of widely- approach, Part II: Handling constraints and distributed weight vectors. Although this extending to an adaptive approach, IEEE algorithm obtains a good approximation of Transactions on Evolutionary Computation the Pareto front (PF) for multi-objective 18 (2014) 602–622. 4. Q. Zhang, H. Li, optimization problems, this may be more MOEA/D: A multiobjective evolutionary difficult to obtain for many-objective algorithm based on decomposition, IEEE optimization problems due to the fact that Transactions on Evolutionary Computation the weight vectors used are never updated 11 (2007) 712–731. 5. Y. Qi, X. Ma, F. Liu, L. along the search process. For this reason, we Jiao, J. Sun, J. Wu, MOEA/D with adaptive propose a new version of the algorithm, weight adjustment, Evolutionary called A-GWASF-GA, in which a dynamic Computation 22 (2014) 231–264. 6. E. adjustment of the weight vectors is carried Zitzler, L. Thiele, M. Laumanns, C. M. out. The main idea is to re-calculate some Fonseca, V. G. da Fonseca, Performance weight vectors in order to obtain solutions in assessment of multiobjective optimizers: an parts of the PF with a lack of solutions. Firstly, analysis and review, IEEE Transactions on a percentage (p) of the total number of Evolutionary Computation 7 (2003) 117– evaluations is performed with the original 132. GWASF-GA [1]. Secondly, during the rest of evaluations (1-p), we re-calculate na times 2. A Reference Set Based the projection directions determined by a Evolutionary Algorithm for Many- subset of Na weight vectors. The re- calculation process is based on a scattering Objective Combinatorial Optimization level, a measure based on the distance of Problems each solution and the solutions around it. *Mert Sahinkoc, Bogazici University, According to the scattering level of the Turkey, [email protected] generated solutions, we detect the Na weight vectors projecting toward overcrowded areas Umit Bilge, Bogazici University, Turkey, of the PF and we re-calculate them so that [email protected]

167

Methods for multi-objective optimization parent selection mechanisms, mutation and frequently suffer scalability issues when local improvement operators. Numerical number of objectives is high. Recently, a experiments demonstrate the success of the number of many-objective evolutionary proposed algorithm compared to some algorithms (MaOEAs) has been proposed to existing approaches over the benchmark characterize and overcome the challenges combinatorial optimization problems. The presented when the number of objectives is findings are inspiring and encouraging in higher than three. MaOEAs are categorized terms of adapting reference set based into several classes based on their key idea techniques to many-objective combinatorial used and reference set based approaches optimization problems. are one of the most promising methods. In these approaches, a set of reference points 3. Information-Based Evolutionary is used to guide the search process and Algorithm for Interactive measure the quality of solutions. Reference Computational Design set based methods differ from one another with respect to how they construct the *Michael S. Bittermann, Istanbul reference set and how they measure the Technical University, Turkey, quality of the solutions during the evolution. [email protected] It is crucial to develop an effective Ozer Ciftcioglu, Delft University of positioning and adaption strategy for the Technology, Netherlands, reference set. It is argued that an adaptive [email protected] reference set strategy is necessary to deal with problems with objectives of different A novel approach to selection in evolutionary scales or problems having an irregular, algorithms is described. It is referred to as degenerate, discontinuous, nonlinear Pareto information-based selection, since the fronts. Briefly, a good quality reference set probability of reproduction of a population should be less sensitive to the problem type member is determined by the amount of and Pareto front geometry and should be information the individual represents in the scalable to many-objective problems. Most of context of the search. The objective is to the existing work on reference set based develop an algorithm for interactive MaOEAs study well-defined continuous computational design. In such applications, mathematical functions with designed identification of optimal solutions is only one Pareto front characteristics and studies on of two motivations, as maximizing problem classes, such as combinatorial information about the designer’s tacit goals optimization, are still rare. In this study, we is the second one. This is in contrast to usual choose the many-objective versions of application of evolutionary algorithm for knapsack problem, quadratic assignment solving engineering or decision-making problem and minimum spanning tree problems, as objective functions are usually problem as the benchmarks. Our proposed given in advance in these problems. The MaOEA combines the effective features of method proposed in this work is to measure some existing many-objective evolutionary the amount of information a population approaches together with several other member carries in two respects, and to fuse prominent evolutionary strategies in an the two measurements thereafter. The first innovative fashion. It uses elitist non- information measurement is based on a dominated sorting and reference points that probabilistic model of the convergence are mapped on a hyperplane. The behaviour of the algorithm. What is hyperplane is constructed at the beginning of measured is how informative an individual is the algorithm by using the lexicographic for reaching optimality. The second optimal solutions and its position in the information measurement is based on a objective space remains fixed over the probabilistic model of the diversity course of the algorithm. An adaptive diminishment during the search. What is reference point strategy is implemented measured is how informative an individual is during the evolutionary search based on for preserving the diversity of the population. gathered information from the population. When an individual is found that is Our algorithm also employs a non-dominated informative in both respects, then this solution archive, a path relinking member is to deem precious in the context of recombination scheme, complementing interactive evolution. The preciousness is

168 quantified using logic operation of the fuzzy FRI-2-F neural tree methodology, combining the two information measurement results. One notes Contributed Session: Making Decisions that the preciousness measure applies for with Environmental Concerns both, single objective and multi-objective Friday 13:00-14:40 - Room: Basilica Cistern design problems. In both cases the diversity preservation aspect is computed in the same Chair: Nahid Rezaeinia manner, as we are concerned with diversity in the decision variable space; and the 1. Assessment of Energy and information fusion is also accomplished via Emission Reducing Measures in the same computations in both cases. The Container Terminals Using difference between single- and multi- PROMETHEE objective cases is with respect to the optimality aspect. In the single-objective *Erik Pohl, Universität Duisburg-Essen, case, optimality is measured by a scalar Germany, [email protected] number. Thus modelling evolutionary Christina Scharpenberg, University convergence refers to probability distribution Duisburg-Essen, Germany, associated with this dimension. In the multi- [email protected] objective case optimality is measured by Jutta Geldermann, University Duisburg- non-dominance in the form of rank or Essen, Germany, jutta.geldermann@uni- strength, depending on the algorithm. Thus modelling the convergence refers to due.de probability distribution associated with the While serving as an interface of sea and land- rank or strength dimension. The side, container terminals play an important characteristics of the novel method are role in global supply chains. The operators of verified for two multi-objective optimization container terminals, on the one hand, are problems. The first one is a known test facing pressure on maintaining or even problem, while the second one is from the getting a higher level of throughput by architectural design domain. For both shipping companies, on the other hand, port- problems the information-based evolutionary authorities and governments put pressure on algorithm is compared against the same energy saving and emission reduction. algorithm implementing tournament Therefore, the goal of many container selection. The experimental results show terminals today is to reduce energy that the information-based approach consumption while maintaining or improving maintains a balance between diversity the current service level. Since most of the preservation in the decision variable domain energy of container terminals is needed for and selection pressure in the objective handling equipment, emissions can be function domain, whereas the conventional reduced by using clean energy or by algorithm sooner exhausts this diversity in replacing old equipment with energy-saving favour of higher selection pressure. The gain devices. The project “Simulation-based in chromosome diversity by the novel evaluation of measures for the improvement algorithm comes at a price, namely inferior of energy sustainability in port operations” convergence, as this is hinted by the ‘no-free- (SuStEnergyPort), carried out by the lunch’ theorem. The increased diversity in University Duisburg-Essen and the Hamburg the decision variable space indicates the Port Consulting GmbH (HPC) aims at suitability of the novel method for interactive developing a structured, model-based computational design. A generic contribution methodology to identify suitable measures of the work to multi-objective evolutionary that port operators can use to improve their optimisation is introduction of information energy efficiency and utilization of renewable fusion during selection as alternative to energy. In this project, a selection of applying a fixed hierarchy among selection promising measures for the abatement of criteria. This work has been accomplished CO2 emissions is implemented in a under the auspice of ITU BAP (Scientific simulation tool covering both logistic and Research Projects Coordination Unit of energetic aspects. As some measures Istanbul Technical University); grant no. MGA- include using new handling equipment, e.g. 2018-41526. The support is gratefully straddle carrier, a Life Cycle Assessment acknowledged. (LCA) is used to measure and compare the

169 environmental impact. In this talk, we considered in solar warehouse location analyze and evaluate different energy selection decisions since the strategic consumption and emission reduction location of a solar warehouse is of vital measures applicable to container terminals importance for the energy efficiency and using PROMETHEE. The assessment is effectiveness. In addition to their based on data derived from the projects case identification, the prioritisation of these study, the container terminal Tollerort (CTT) criteria also needs to be evaluated in order in Hamburg, and the simulation tool. We to provide a realistic and powerful solution consider ecological, economic and social on reducing the environment pollution. criteria and analyze the trade-off between Thus, in order to deal with these problems, energy or emission saving and service levels. this study first aims to determine the solar The SIMOS method is used to determine the criteria and their importance ratings for solar weights of the criteria for multiple warehouses. Furthermore, it is also aimed to stakeholders of the container terminals, e.g. examine how these solar criteria affect operators and port-authorities. On the basis strategic location selection decisions of solar of this assessment, a roadmap towards a warehouses. Accordingly, in this study, the more profitable and ecological port operation solar warehouse location selection criteria can be developed. and their ratings are determined through the Analytical Hierarchy Process (AHP), Analytical 2. Strategic Location Selection for Network Process (ANP), and Techniques for Solar Warehouses: The Green Order Preference by Similarity to Ideal Logistics Perspective Solution (TOPSIS) methods. More specifically, the AHP is used to examine the *Rasih Boztepe, Reysaş Taşımacılık ve importance of solar criteria for selecting the Lojistik A.Ş., Turkey, [email protected] solar warehouse location. The first three Onur Cetin, Trakya university, Turkey, criteria appeared from the findings, which [email protected] have a weight of 67% in total, were added to Berk Kucukaltan, Trakya University, the existing warehouse location selection Turkey, [email protected] criteria for green logistics as this leads to the Necdet Özçakar, Istanbul University, second part of the application performed by Turkey, [email protected] the ANP. Finally, the outcomes obtained from the first two methods formed the decision Environmental pollution that results from an matrix to be used in the TOPSIS technique, increase in greenhouse gas emission is one which concluded the best solar warehouse of the major problems that humankind has location among five province alternatives. At been encountered. In order to deal with the end of these stages, the findings of this these problems, several approaches are study show that İzmir has the highest order largely experienced in the supply chain of preference, with the value of 0.61, among domain and one the most prominent creative province alternatives, followed by Istanbul solutions shown in this context is the (0.58), Adana (0.51), Ankara (0.46), and extensive use of green warehouses, which Samsun (0.36). To conclude, the proposed are principally solar warehouses that model holds novelties in the establishment generate electricity by solar energy panels on of solar warehouse selection criteria and in the roof tops. In practice, implementation of the prioritisation of solar warehouses. While the solar warehouses is regarded as a new doing these, the application of multiple approach not only for the sustainable MCDM techniques in this strategic decision logistics concept, but also for advancing plays a significant role to contribute to the environmentally sensitive green logistics. literature. Moreover, the presented findings Yet, while implementing solar warehouses as also contribute to the advancement of a significant approach in the supply chain research on solar warehouse selection in the domain, more particularly in the logistics green logistics literature and to the practice area, the selection of a warehouse location related to strategic decision-making on remains as a multi criteria decision making location selection for solar warehouses. (MCDM) problem, where numerous Keywords: Analytic Hierarchy Process (AHP), quantitative and qualitative criteria exist. In Analytic Networking Process (ANP), such abundancy, it becomes essential to Technique for Order Preference by Similarity identify which solar criteria should be to Ideal Solution (TOPSIS), Strategic

170

Decision-Making, Green Logistics, Solar preferred alternative. In the beginning (until Warehouse. the mid-1980s), only one criterion was considered related to the minimization of the 3. Multicriteria Methods and the energy generation cost (construction, Hydropower Plants Planning in Brazil operation and maintenance). From then, environmental issues became more relevant *Igor Raupp, Electric Energy Research and a multicriteria decision process was Center - CEPEL, Brazil, [email protected] carried out, incorporating in the 1990s the Fernanda Costa, Electric Energy minimization of negative socio- Research Center - CEPEL and Rio de environmental impacts and, later, in the 21st Janeiro State University - UERJ, Brazil, century, the maximization of positive socio- [email protected] environmental impacts, the less significant João Clímaco, INESC - University of criterion. Nowadays, to select the most Coimbra, Portugal, [email protected] preferred alternative, the three criteria are Marcelo Miguez, COPPE/UFRJ, Brazil, aggregated into a single index using the [email protected] weighted sum method, and the alternative with the best value is selected. So, the Over the years, Brazil has been taking multidisciplinary team conducting the study advantage of its hydroelectric potential in must define specific weights for each order to make the country self-sufficient in criterion, reflecting the opinion of the society. electricity, based on a clean, renewable and The choice of weights is a subjective task, it low-cost source. The hydroelectric potential is difficult to quantify the decision makers’ of Brazilian river basins (approximately preferences, it may change during the 250,000 MW) is being progressively decision process and in a group decision, the exploited since the beginning of the last opinions and preferences can often diverge. century, and today, with 44% of the potential Aiming the continuous improvement of the already exploited, this type of electricity decision making of the Inventory Studies, in generation represents 65% of the Brazilian this article two multicriteria methods are electric power matrix, with more than 200 tested in order to mitigate this difficulty, hydroelectric plants with capacity greater avoiding to ask precise values for the than 30 MW in operation. As a major user of weights, but still allowing to distinguish the water, the electricity sector has the importance of the criteria. Applications have responsibility and duty to plan the use of this shown that the VIP Analysis and Electre III resource as an input for the production of methods are very promising and strong electricity in a rational and optimized way, candidates. The paper details the three alongside the other users of water in the river criteria, compares the results of the weighted basin. The planning of the hydroelectric sum with those of these two methods and expansion in Brazil is done through a series discusses how the parameters and other of studies that consider different time information demanded by these methods horizons and successive approximations. should be defined. For this, the methods One of the first steps to the development of were applied in a real Inventory Study of a hydropower plants is the River Basin Brazilian river basin. Lastly, expanding Inventory Studies and its main objective is to energy generation means addressing a design and analyze all possible alternatives paradox in society: people want more energy of group of hydropower plants in a river basin for development, but they also question how in order to select the most preferred one, it is produced and how it affects the which will determine the hydroelectric environment. Good planning and transparent potential to be exploited in the basin. After decision-making processes are the way to this stage, each hydropower plant that resolve such conflicts. Thus, improvements composes the alternative selected in the in strategic decision making, such as the one Inventory Studies will be studied separately in Inventory Studies, should always be to analyze its feasibility in detail. The pursued. methodologies of the Inventory Studies have undergone several updates, mainly concerning the decision making, incorporating the yearnings of the society so that this choice effectively reflects the most

171

4. Evaluating a Forest Road Network attributes in order to design efficient forest by an Additive Value Model road network (Hayati et al., 2013). In this research, we propose a multi-criteria method *Nahid Rezaeinia, Norwegian School of in order to evaluate the existing road network Economics, Norway, in Caspian forest in Iran. We take into [email protected] account quantitative and qualitative Mario Guajardo, Norwegian School of attributes such as road length, road width, Economics, Norway, road safety, ease of road construction, ease [email protected] of progression to the mountainous area, less destruction of wildlife habitats and Road planning and development is an environmental features, ease of access to essential component in forest management. the road for villagers. The aim is to find a It improves operational forest harvesting weight setting that reflects each attribute’s through providing better access to timber importance in relationship to other harvest sites and availability of product attributes. UTA (UTilités Additives) method market for rural population. In forest road based on an additive value model is planning, it is important to address safety of proposed for evaluation. The method takes a forest products transport, comfort and given ranking for a set of alternatives, which economy of vehicle operations. Caspian in this work consists of the existing road forests (i.e., forests in north Iran) that are alternatives. The procedure of the method is located between north of Alborz Mountains inferring additive value functions to the and south of Caspian Sea are the only alternative set. The method uses linear commercial forests of Iran. Because of high- programming to assess functions and to find quality timbers, access to this area has weights for attributes so that the rankings industrial importance for wood industry obtained through these functions on activists. In addition, the environmental alternative set is as consistent as possible protection of this area is important for forest with the given one. To the best of our management in Iran. Hence, for this area, knowledge, this is the first study that uses having an efficient planning of road network the UTA method in the field of forest that incorporates economic and management. According to our experience environmental considerations is critical. A UTA method, works well for this application. great body of literature addresses the use of This method is flexible in terms of adding or multi-criteria decision-making methods for changing attributes. Furthermore, due to planning and evaluating various aspects of elimination of some time-consuming forest management. Analytic Network process, such as pairwise comparisons, the Process (ANP) has been utilized for use of this method is easy in comparison of evaluating performance of sustainable some other methods such as AHP and ANP. management strategies (Wolfslehner and Keywords: Forest roads, Multi-criteria Vacik, 2008). Naghdi and Babapour (2009) decision making, Additive value model. proposed an Analytical Hierarchy Process (AHP) for evaluating and classifying the FRI-3-L effective factors in forest road construction. Closing Session A hybrid MCDM method has been developed for evaluating attributes and alternatives in Friday 14:40-15:10 - Room: Galata Tower forest roads locating (Hashemkhani et al, Chair: Ilker Topcu 2011). Delphi and AHP method have been proposed for selecting important attributes and obtaining relative importance of

172

Author Index

A Atan, Murat 138 Atıcı, Bengü 145 Abdelkader, Mendas 35 Aydın, Umut 89 Abdellah, Mebrek 35 Äyrämö, Sami 156 Abushammala, Mohammed 116 Azad, Nader 139 Adebiyi, Sulaimon 72 Azizi, Majid 80 Adıyeke, Esra 38 Afsar, Bekir 103 B Agell, Núria 81 Babashov, Vusal 68 Aghaei Pour, Pouya 102 Baburoglu, Oguz 37 Aguirre, Eduar 53 Bahiense, Laura 60, 61 Ahonen-Jonnarth, Ulla 149 Barbati, Maria 42 Ak, Cem 123 Bashir, Bashir 64 Akhtar, Khalid 58 Başar, Ayfer 161 Aktaş, Dilay 77 Bayar, Mustafa 138 Al-Rahbia, Alia 116 Bayazit, Ozden 33 Alberti, Alexandre 49 Baydoğan, Mustafa 38 Alptekin, Can 93 Baylan, Emin Başar 97 Altug, Gorkem 43 Bayram, Gökçen 84 Alves, Maria João 79, 104 Beauregard, Yvan 69 Amami, Dalel 83 Belahcene, Khaled 159 Amira Nagham, Mokrani 91 Beldek, Tuğçe 42 Andersson, Hanna 149 Ben Amor, Sarah 68, 69 Andreolli, Francesca 100 Berk, Erhan 133 Anisic, Zoran 114 Bilge, Umit 167 Arcidiacono, Sally Giuseppe 130 Bişkin, Büşra 107 Arias, Adriana 47 Bittermann, Michael S. 168 Armas, Iuliana 53 Bohanec, Marko 98 Arslan, Ozan 123 Borbinha, José 166 Asan, Umut 113, 136 Boujelben, Mohamed Aymen 110 Aspen, Dina 35 Boujelbene, Younes 40 Atac, Esra 59 Boumesbah, Asma 67

173

Bouyssou, Denis 112 Cinaroglu, Songul 90 Boztepe, Rasih 170 Civelek, Merve 144 Bökler, Fritz 126 Clímaco, João 69, 171 Bökman, Fredrik 149 Cooper, Orrin 31, 34, 49, 62 Bravo, Juan 47 Correa-Espinal, Alexander 64 Bregar, Andrej 55 Corrente, Salvatore 130, 166 Brunelli, Matteo 30, 75 Cortes Aldana, Felix Antonio 46 Buyukozkan, Gulcin 162 Cortés, Pablo 64 Büyüközkan, Gülçin 82 Costa, Ana Paula 88 C, Ç Costa, Ana Sara 166 Costa, Fernanda 171 Cabello, José M. 140 Cvetkoska, Violeta 147 Cabrera-Guerrero, Guillermo 141 Cagdas, Gulen 150, 151 D Cakir, Sercan 86 D'Alpaos, Chiara 100 Camgöz Akdağ, Hatice 42 Daechert, Kerstin 63 Campo, Emiro 64 Daher, Suzana 88 Cano, Jose 64 Dakare, Olamitunji 72 Caprace, Jean-David 60, 61 Dammak, Fatma 83 Caprila, Alexandra 45 Dangol, Ramesh 32 Cavalcante, Crıstıano Alexandre 49 Darende, Pinar 117 Cavallo, Bice 75 Dasdemir, Erdi 108 Çaybaşı, Gökhan 162 De Arruda, Edilson 61, 62 Cayir Ervural, Beyzanur 99 De Smet, Yves Cebesoy, Melodi 159 55, 110, 111, 132, 165 Cenani, Sehnaz 150, 151 De Souza, Marcelo Igor 60, 61 Cerit, Bülent 124 Decaestecker, Christine 132 Cetin, Onur 170 Dehghani, Mohammad 85 Çetinkaya, Ömer Mücahid 123 Dehnokhalaji, Akram 89, 97 Ceyhan, Gokhan 78 Dejaegere, Gilles 110 Chakhar, Salem 110 Delen, Dursun 99 Chen, Yu-Wang 73 Demir, Ezgi 59 Chergui, Mohamed El-Amine 67 Deniz, Nurcan 135 Chugh, Tinkle 91 Dhurkari, Ram 146 Ciftcioglu, Ozer 168 Djamal, Chaabane 40, 91

174

Djeffal, El Amir 106 Ghazli, Kahina 51 Doan Nguyen, Anh Vu 165 Ghorbel, Amira 40 Doganay, Onur Tanil 92 Gillis, Nicolas 51 Doubrava, Vít 92 Gómez, Trinidad 140 Doumpos, Michalis 123 Gonzalez Gallardo, Sandra 167 Duleba, Szabolcs 100 González-Urango, Hannia 72 Duzdar Argun, Irem 58 Gonzalez-Urango, Hannia 84 E Gottschalk, Hanno 92, 93 Greco, Salvatore 42, 56, 74, 131, 166 Egdemir, Emine 133 Groselj, Petra 145 Ehrgott, Matthias 142 Guajardo, Mario 172 Eisenstadt-Matalon, Erella 153 Guerreiro, Andreia P. 128 El Gibari, Samira 140 Gurbuz, Mustafa Zahıd 119 Engau, Alexander 157 Gutierrez, Maicholl 141 Eren, Tamer 132 Gül, Sait 162, 163 Erol, Eren 137 Güleç, Nurullah 35 F Günhan, Büşra 70

Faleschini, Flora 100 H Fedrizzi, Michele 45 Hakanen, Jussi 102, 117, 156 Ferreira, Rodrigo 87, 88 Hämäläinen, Raimo 134 Figueira, José Rui 42, 50, 55, 56, 166 Hamurcu, Mustafa 132 Figueiredo, Ciro 49 Harale, Nitin 44 Filho, Angelo 118 Hayashida, Tomohiro 65, 119, 164 Filiz Türkmen, Gizem 148 Hecht, Robert 86 Fleming, James 45 Henggeler Antunes, Carlos 79, 104 Frini, Anissa 110, 131 Hernandez Castellanos, Carlos 51 G Hodgett, Richard 74 Hsu, Ching-Chun 116 Garcia-Melon, Monica 46, 72, 85 Huang, Chi-Yo 116 Gatica, Gustavo 141 Huang, He 165 Gavris, Alexandru 53 Huang, Shan-Lin 71 Geldermann, Jutta 169 Hubinont, Jean-Philippe 55 Genc, Tolga 29 Ghaderi, Mohammad 127

175

I, İ Khurram Ali, Hafiz 58 Kılıç, Hakan 90 Infante, Eduardo 60, 61 Kik, David 52 İnkaya, Tülin 77 Kilani, Arij 40 Ishizaka, Alessio 42, 110 Kirlik, Gökhan 39 Isik, Mine 99 Kizil, Kerim Uygur 59 J Klamroth, Kathrin 50, 81, 92 Koçoğlu, Fırat 123 Jabbarzadeh, Armin 139 Koker, Derya 37 Jahanzaib, Mirza 58 Koksal, Gulser 76 Jiménez-Martín, Antonio 66 Konur, Dincer 130 Johansen, Børge Andreas 35 Konyalioglu, Kemal 42 Juszczuk, Przemysław 120, 121, 142 Korhonen, Pekka 29, 89, 97, 104 K Korkut, Kübra 124 Köksalan, Murat Kabadayi, Nihan 85 76, 78, 107, 108, 109, 160 Kabak, Özgür 35, 36, 90, 125 Krasny, Pawel 41 Kadaifci, Cigdem 43 Krátký, Tomáš 91 Kaliszewski, Ignacy Krellenberg, Kerstin 86 106, 120, 121, 142 Kucukali, Ozdemir 125 Kanazawa, Yuuki 65 Kucukaltan, Berk 170 Karabay, Nail 109 Kułakowski, Konrad 46 Karadayı, Melis Almula 89, 137 Kumar Mangla, Sachin 115 Karakaya, Gülşah 160 Karasakal, Esra 144, 145 L

Karasakal, Orhan 145 Labib, Ashraf 110 Kargar, Mehdi 75 Lage, Carla 87 Karpak, Birsen 32 Lagos, Carolina 141 Karsu, Özlem 64 Lalic, Bojan 114 Katsigiannis, Fotios 137 Lannez, Sebastien 155 Kazançoğlu, Yiğit 115 Li, Xia 71 Kerbérénès, Antoine 152 Liang, Fuqi 30 Keskin, İrem Nur 63 Lokman, Banu 77, 78 Keskinocak, Pınar 29 López-Ibáñez, Manuel 138 Khanıyev, Tahir 117 Lundy, Michele 74

176

Luo, Jin-Hua 71 Moulai, Mustapha 51 Luque, Mariano 167 Moumni, Nourchen 166 M Mousavi, Seyed Mohsen 101 Mousseau, Vincent 110, 159 Macharis, Cathy 165 Mu, Enrique 31 Macuada, Claudio 57 Mukhametzyanov, Irik 96 Makranfar, Zohra 94 Mansikka, Heikki 134 N Mantogiannis, Vasileios 137 Nasrabadi, Nasim 89, 97 Manyoma, Pablo 47, 86 Negüs, Fahri 82 Marchant, Thierry 112 Neyra Belderrain, Mischel Carmen 88 Marjanović, Ivana 83 Nishizaki, Ichiro 65, 119, 164 Marjanovic, Ugljesa 114 Nohadani, Omid 142 Marsaro, Mônica 49 Nowak, Maciej 101 Martins, Isabelle 60, 61 O Mateos, Alfonso 66 Mebrek, Abdellah 94 Ober-Blöbaum, Sina 51 Mechitov, Alexander 29, 48 Oddershede, Astrıd 57 Medic, Nenad 114 Ojalehto, Vesa 156 Medrano, F. Antonio 152 Olhofer, Markus 102 Miclea, Adriana 44 Omoera, Charles 72 Miettinen, Kaisa Onan, Kıvanc 119 91, 101, 102, 103, 117, 138 Ono, Masato 164 Miguez, Marcelo 171 Ouerdane, Wassila 159 Miroforidis, Janusz 106, 120, 121, 142 Ouznadji, Said 40 Moncef, Abbas 39 Oyenuga, Olamilekan 72 Montazer, Gholamali 113 Ozpeynirci, Ozgur 143, 144 Moraes, Fernanda 60, 61 Oztek, Yaman 82 Morales, Daniela 47 Ozturk, Onur (Çimtaş) 112 Moralioglu, Begum 150 Ozturk, Onur (uOttawa) 139 Moshaiov, Amiram 153 Ö Moshkovich, Elena 29, 48 Moslem, Sarbast 100 Önsel Ekici, Şule 125 Mota, Caroline 49 Özarslan, Ali 160 Mouhib, Youness 131 Özateş Gürbüz, Melis 76

177

Özaydın, Özay 54, 99 R Özçakar, Necdet 170 Özdemir, Mustafa 133 Raupp, Igor 171 Özgen, İrem 113 Rebaï, Abdelwaheb 166 Özkan Özen, Yeşim Deniz 115 Reinhardt, Gilles 68 Özkardeş, Sibel 95 Rezaei, Jafar 30, 31 Özmez, Ali 93 Rezaeinia, Nahid 172 Özpeynirci, Selin 143, 144 Rodemann, Tobias 102 Rodrigues, Márcio 57 P Romero, Jorge 46 Palominos, Pedro 58 Rosenfeld, Jean 132 Panaro, Simona 42 Roumeissa, Kerbouı 39 Paquet, Marc 69 Rouyendegh, Babak 117 Paquete, Luís 50 Roy, Bernard 56 Pavlou, Antonis 123 Ruiz, Ana B. 167 Peasley, Michael 31 Ruiz, Francisco 103, 140 Peker, İskender 162 S Pelissari, Renata 68 Pereyra-Rojas, Milagros 31 Saarela, Mirka 156 Pérez-Gladish, Blanca 72 Saborido, Rubén 167 Pessatto Mondadori, Jorge Augusto 88 Sagir, Mujgan 133 Pessôa, Leonardo Antonio 87 Şahin Zorluoğlu, Özge 36 Piercy, Craig 154 Sahinkoc, Mert 167 Pirlot, Marc 112, 159 Saidi, Ahmed 94 Planellas, Marcel 81 Saini, Bhupinder 117 Podkopaev, Dmitry Salah, Mohamed Essalah 40 101, 120, 121, 142 Samanlioglu, Funda 123 Pohl, Erik 169 Sanchez, Monica 81 Polat, Seçkin 113 Sandholm, Tuomas 126 Popović, Žarko 83 Santiago, Katarina 88 Popovici, Diana 53 Savaş, Sezer 151 Porro, Olga 81 Sayın, Serpil 39 Q Scharpenberg, Christina 169 Qazi, Wajeeha 116 Schultes, Johanna 92 Quezada, Luis 58 Schulze, Britta 50, 63

178

Sekizaki, Shinya 65, 119, 164 Trzaskalik, Tadeusz 101 Shavazipour, Babooshka 138 Tuncer Şakar, Ceren 107, 159 Shayganmehr, Masoud 113 Tureci, Hannan 107 Shentall, Glyn 142 Tzeng, Gwo-Hshiung 71 Shibukawa, Naoyuki 65 U, Ü Sınayiş, Nihat Can 78 Silva, Lucio 88 Ucal Sari, Irem 125 Siraj, Sajid 74 Uenishi, Hirokazu 65 Šírová, Eva 57 Ülengin, Burç 89 Sitarz, Sebastian 101 Ülengin, Füsun 54, 80, 89, 125 Skulimowski, Andrzej 105 Ulus, Firdevs 63 Soares, Henrique 61 Unver, Berna 80, 125 Sobrie, Olivier 159 Uzturk, Deniz 82 Spengler, Thomas S. 52 Uzun Bayar, Irmak 138 Stanković, Jelena 83 V Steuer, Ralph 154 Stewart, Theodor 55 Vanderpooten, Daniel 152 Stiglmayr, Michael 50, 92 Vanfossan, Samuel 130 Stubington, Emma 142 Vanpeperstraete, Jean-Michel 152 Swain, Anjan Kumar 146 Vargas, Luis 143 Szmurło, Robert 142 Verma, Manish 140 Vinagre-Fernandez, Maria Regla 72 T Virtanen, Kai 134 Tasic, Nemanja 114 W Tavora, Giselle 60, 61 Teichert, Katrin 63 Wallenius, Hannele 154 Teixeira de Almeida, Adiel 87, 88 Wallenius, Jyrki 29, 89, 97, 154 Tello, Faustino 66 Wang, Yu-Tai 116 Tezcaner Öztürk, Diclehan Weidner, Petra 51 107, 108, 109 Wichmann, Matthias G. 52 Thameri, Messaoud 40 Wiecek, Margaret 122 Toktas Palut, Peral 119 Wu, Jei-Zheng 43 Toma-Danila, Dragos 53 X Topcu, Ilker 80, 125, 148 Tozan, Hakan 137 Xu, Peng 29

179

Y Z

Yamamoto, Hiroyuki 119 Zaabar, Imen 69 Yamasaki, Takuya 65 Zadnik Stirn, Lidija 145 Yang, Chia-Lee 116 Zaim, Selim 99 Yanık, Seda 78, 94, 95 Zakaria, Sadouni 35 Yanmaz, Ozgur 43, 136 Zavadil, Lukáš 91 Yavuz, Idil 34 Zawadzka, Anna 142 Yet, Barbaros 159 Zhang, Yao 40 Yılmaz, Beyza Özlem 137 Zohra, Mekranfar 35 Yılmaz, Salim 94 Zopounidis, Constantin 123 Yunculer, Kutay 149

180

List of Chairs

Amami, Dalel TUE-3-A (Tuesday 13:50-15:30) Azad, Nader THU-1-E (Thursday 9:00-10:40) Ben Amor, Sarah TUE-1-A (Tuesday 9:00-10:40) Bittermann, Michael S. FRI-2-E (Friday 13:00-14:40) Bohanec, Marco TUE-4-A (Tuesday 16:00-17:40) Brunelli, Mateo TUE-1-C (Tuesday 9:00-10:40) Cakir, Sercan TUE-3-B (Tuesday 13:50-15:30) Cavalcante, Cristiano MON-3-C (Monday 13:50-15:30) Cenani, Sehnaz THU-4-C (Thursday 14:20-16:00) Cooper, Orrin MON-2-B (Monday 11:10-12:50) Costa, Ana Sara FRI-2-C (Friday 13:00-14:40) Daher, Suzana TUE-3-C (Tuesday 13:50-15:30) Darende, Pinar WED-1-C (Wednesday 9:00-10:40) Doumpos, Michalis WED-1-E (Wednesday 9:00-10:40) Duzdar Argun, Irem MON-4-B (Monday 16:00-17:40) Ehrgott, Matthias THU-1-F (Thursday 9:00-10:40) Engau, Alexander THU-4-F (Thursday 14:20-16:00) Ferreira, Rodrigo J P TUE-3-C (Tuesday 13:50-15:30) Figueira, Jose Rui MON-2-F (Monday 11:10-12:50) WED-3-P (Wednesday 13:10-14:25) Garcia-Melon, Monica TUE-1-B (Tuesday 9:00-10:40) Greco, Salvatore MON-4-A (Monday 16:00-17:40) Gül, Sait FRI-2-B (Friday 13:00-14:40) Hakanen, Jussi MON-2-E (Monday 11:10-12:50) Hämäläinen, Raimo THU-1-B (Thursday 9:00-10:40) Harale, Nitin MON-3-A (Monday 13:50-15:30) Hayashida, Tomohiro WED-1-D (Wednesday 9:00-10:40) Kabak, Özgür TUE-3-D (Tuesday 13:50-15:30) TUE-4-D (Tuesday 16:00-17:40) Kaliszewski, Ignacy TUE-4-E (Tuesday 16:00-17:40) Karakaya, Gülşah FRI-2-A (Friday 13:00-14:40) Karasakal, Esra THU-4-A (Thursday 14:20-16:00) Karpak, Birsen MON-3-D (Monday 13:50-15:30) 181

MON-4-D (Monday 16:00-17:40) Karsu, Özlem MON-4-E (Monday 16:00-17:40) Köksalan, Murat MON-1-P (Monday 9:40-10:40) TUE-4-F (Tuesday 16:00-17:40) WED-2-P (Wednesday 11:10-12:10) THU-3-U (Thursday 13:10-14:10) Lokman, Banu TUE-1-E (Tuesday 9:00-10:40) Mantogiannis, Vasileios THU-1-C (Thursday 9:00-10:40) Martins, Isabelle D. MON-4-C (Monday 16:00-17:40) Mateos, Alfonso MON-4-F (Monday 16:00-17:40) Medic, Nenad WED-1-B (Wednesday 9:00-10:40) Miettinen, Kaisa TUE-4-C (Tuesday 16:00-17:40) Moslem, Sarbast TUE-4-B (Tuesday 16:00-17:40) Mota, Caroline MON-3-C (Monday 13:50-15:30) Önsel Ekici, Şule WED-1-F (Wednesday 9:00-10:40) Özaydın, Özay MON-2-D (Monday 11:10-12:50) MON-3-F (Monday 13:50-15:30) Pirlot, Marc WED-1-A (Wednesday 9:00-10:40) Rezaei, Jafar MON-2-A (Monday 11:10-12:50) Rezaeinia, Nahid FRI-2-F (Friday 13:00-14:40) Romero, Jorge MON-3-B (Monday 13:50-15:30) Rosenfeld, Jean THU-1-A (Thursday 9:00-10:40) Sayın, Serpil TUE-2-P (Tuesday 11:10-12:10) Steuer, Ralph THU-4-E (Thursday 14:20-16:00) Stiglmayr, Michael TUE-3-E (Tuesday 13:50-15:30) Şahin Zorluoğlu, Özge MON-2-C (Monday 11:10-12:50) Tezcaner Öztürk, Diclehan TUE-4-F (Tuesday 16:00-17:40) Topcu, Ilker MON-0-O (Monday 9:00-9:40) THU-2-P (Thursday 11:10-12:10) THU-4-B (Thursday 14:20-16:00) FRI-3-L (Friday 14:40-15:10) Ulengin, Fusun TUE-1-F (Tuesday 9:00-10:40) Ulus, Firdevs MON-4-E (Monday 16:00-17:40) Wallenius, Jyrki FRI-1-M (Friday 9:30-12:00) Weidner, Petra MON-3-E (Monday 13:50-15:30) Yanık, Seda TUE-3-F (Tuesday 13:50-15:30)

182

List of Participants

Abdelkader, Mendas Andreolli, Francesca Centre des Techniques Spatiales, DICEA - University of Padova, Italy Algeria [email protected] [email protected] Arcidiacono, Sally Giuseppe Adıyeke, Esra University of Catania, Italy Boğaziçi University, Turkey [email protected] [email protected] Arias, Adriana Agell, Núria Universidad del Valle, Colombia ESADE-URL, Spain [email protected] [email protected] Asan, Umut Aghaei Pour, Pouya Istanbul Technical University, Turkey University of Jyvaskyla, Finland [email protected] [email protected] Aspen, Dina Aguirre, Eduar NTNU, Norway Universidad del Valle, Colombia [email protected] [email protected] Azad, Nader Aktaş, Dilay University of Ontario Institute of Middle East Technical University, Turkey Technology, Canada [email protected] [email protected] Alves, Maria João Azizi, Majid University of Coimbra, Portugal University of Tehran, Iran [email protected] [email protected] Amami, Dalel Baburoglu, Oğuz University of Sfax, Tunisia Arama Search, Sabanci University, [email protected] Turkey Amira Nagham, Mokrani [email protected] USTHB, Algeria Baskak, Murat [email protected] Istanbul Technical University, Turkey Andersson, Hanna [email protected] University of Gävle, Sweden Başar, Ayfer [email protected] Istanbul Technical University, Turkey [email protected]

183

Bayar, Mustafa Boztepe, Rasih Ankara Haci Bayram Veli University, Reysaş Taşımacılık ve Lojistik A.Ş., Turkey Turkey [email protected] [email protected] Bayazit, Ozden Bökler, Fritz CWU, United States Osnabrück University, Germany [email protected] [email protected] Baylan, Emin Başar Bökman, Fredrik Istanbul Ticaret University, Turkey University of Gävle, Sweden [email protected] [email protected] Bayram, Gökçen Bregar, Andrej Marmara University, Turkey Informatika d.d., Slovenia [email protected] [email protected] Ben Amor, Sarah Brunelli, Matteo University of Ottawa, Canada University of Trento, Italy [email protected] [email protected] Berk, Erhan Büyüközkan, Gülçin National Defense University, Turkey Galatasaray University, Turkey [email protected] [email protected] Bişkin, Büşra Cabrera-Guerrero, Guillermo Gazi University, Turkey Pontificia Universidad Católica de [email protected] Valparaíso, Chile Bittermann, Michael S. [email protected] Istanbul Technical University, Turkey Cakir, Sercan [email protected] Leibniz Institute of Ecological Urban and Bohanec, Marko Regional Development, Germany Jožef Stefan Institute, Slovenia [email protected] [email protected] Camgöz Akdağ, Hatice Boumesbah, Asma Istanbul Technical University, Turkey USTHB, Algeria [email protected] [email protected] Cano, Jose Boumesbah, Asma Universidad de Medellín, Colombia USTHB, Algeria [email protected] [email protected]

184

Cavalcante, Cristiano Costa, Ana Sara Universidade Federal de Pernambuco, Universidade de Lisboa, Portugal Brazil [email protected] [email protected] Cvetkoska, Violeta Cayir Ervural, Beyzanur Ss.Cyril and Methodius University in Konya Food and Agriculture University, Skopje, North Macedonia Turkey [email protected] [email protected] Çevik Onar, Sezi Cebesoy, Melodi Istanbul Technical University, Turkey Hacettepe University, Turkey [email protected] [email protected] Daechert, Kerstin Cenani, Sehnaz Fraunhofer ITWM, Germany Istanbul Technical University, Turkey [email protected] [email protected] Daher, Suzana Cerit, Bülent Universidade Federal de Pernambuco, Istanbul Technical University, Turkey Brazil [email protected] [email protected] Chakhar, Salem Darende, Pinar University of Portamouth, United TOBB ETU, Turkey Kingdom [email protected] [email protected] Dasdemir, Erdi Chen, Yu-Wang Hacettepe University, Turkey The University of Manchester, United [email protected] Kingdom De Smet, Yves [email protected] Université libre de Bruxelles, Belgium Cinaroglu, Songul [email protected] Hacettepe University, Turkey Dehghani, Mohammad [email protected] Northeastern University, United States Civelek, Merve [email protected] ASELSAN, Turkey Dehnokhalaji, Akram [email protected] Aston University, United Kingdom Cooper, Orrin [email protected] University of Memphis, United States Dejaegere, Gilles [email protected] Université libre de Bruxelles, Belgium [email protected]

185

Demir, Ezgi Ervural, Bilal Piri Reis University, Turkey Necmetin Erbakan University, Turkey [email protected] [email protected] Deniz, Nurcan Fedrizzi, Michele Eskişehir Osmangazi Univesity, Turkey University of Trento, Italy [email protected] [email protected] Dhurkari, Ram Ferreira, Rodrigo IFMR Graduate School of Business CDSID/UFPE, Brazil Sricity, India [email protected] [email protected] Figueira, José Rui Djeffal, El Amir Technical University of Lisbon, Portugal University of Batna, Algeria [email protected] [email protected] Filho, Angelo Doganay, Onur Tanil UTFPR, Brazil University of Wuppertal, Germany [email protected] [email protected] Filho, Angelo Doumpos, Michalis UTFPR, Brazil Technical University of Crete, Greece [email protected] [email protected] Fonseca, Carlos M. Duzdar Argun, Irem INESC Coimbra - Universidade de Duzce University, Turkey Coimbra, Portugal [email protected] [email protected] Ehrgott, Matthias Frini, Anissa Lancaster University, United Kingdom Université du Québec à Rimouski, [email protected] Canada Eisenstadt-Matalon, Erella [email protected] ORT Braude Colleage, Israel Garcia-Melon, Monica [email protected] Universitat Politecnica de Valencia, El Gibari, Samira Spain University of Málaga, Spain [email protected] [email protected] Geyik, Cem Engau, Alexander Çimtaş, Turkey Dalhousie University, Canada [email protected] [email protected]

186

Ghaderi, Mohammad Hakanen, Jussi Pompeu Fabra University (UPF), Spain University of Jyvaskyla, Finland [email protected] [email protected] Ghazli, Kahina Hamadou, Sara USTHB, Algeria USTHB, Algeria [email protected] [email protected] Gonzalez Gallardo, Sandra Hämäläinen, Raimo University of Malaga, Spain Aalto University, Finland [email protected] [email protected] Gonzalez-Urango, Hannia Hamurcu, Mustafa INGENIO (CSIC-UPV), Spain Kirikkale University, Turkey [email protected] [email protected] Gök, Filiz Harale, Nitin Dogus University, Turkey Ecole centrale de Lille, France [email protected] [email protected] Greco, Salvatore Hayashida, Tomohiro University of Catania, Italy Hiroshima University, Japan [email protected] [email protected] Groselj, Petra Henggeler Antunes, Carlos University of Ljubljana, Slovenia University of Coimbra, Portugal [email protected] [email protected] Guerreiro, Andreia P. Hernandez Castellanos, Carlos University of Coimbra, Portugal University of Oxford, United Kingdom [email protected] [email protected] Gül, Sait Hodgett, Richard Beykent University, Turkey University of Leeds, United Kingdom [email protected] [email protected] Güleç, Nurullah Huang, He Ankara Yıldırım Beyazıt University, Vrije Universiteit Brussel, Belgium Turkey [email protected] [email protected] Huang, Shan-Lin Günhan, Büşra Sanming University, China Kütahya Dumlupınar University, Turkey [email protected] [email protected]

187

Huang, Chi-Yo Karasakal, Esra Taiwan Normal University, Taiwan Middle East Technical University, Turkey [email protected] [email protected] Hubinont, Jean-Philippe Karasakal, Orhan Université Libre de Bruxelles, Belgium Çankaya University, Turkey [email protected] [email protected] Işık, Mine Kargar, Mehdi Boğaziçi University, Turkey Ryerson University, Canada [email protected] [email protected] İmamoğlu, Gül Karpak, Birsen Istanbul Technical University, Turkey Youngstown State University, United [email protected] States Johansen, Børge Andreas [email protected] NTNU, Norway Karsu, Özlem [email protected] Bilkent University, Turkey Juszczuk, Przemysław [email protected] University of Economics, Katowice, Kazançoğlu, Yiğit Poland Yaşar University, Turkey [email protected] [email protected] Kabadayı, Nihan Kerbérénès, Antoine Istanbul University, Turkey Université Paris Dauphine, France [email protected] [email protected] Kabak, Özgür Keskinocak, Pınar Istanbul Technical University, Turkey Georgia Tech, United States [email protected] [email protected] Kadaifci, Cigdem Khurram Ali, Hafiz Doğuş University, Turkey Unviversity of Engineering and [email protected] Technology, Pakistan Kaliszewski, Ignacy [email protected] Polish Academy of Sciences, Poland Kılıç, Hakan [email protected] Koç University, Turkey Karakaya, Gülşah [email protected] Middle East Technical University, Turkey Kik, David [email protected] TU Braunschweig, Germany [email protected]

188

Kizil, Kerim Uygur Lannez, Sebastien Istanbul Technical University, Turkey FICO Xpress Optimization, United [email protected] Kingdom Klamroth, Kathrin [email protected] University of Wuppertal, Germany Liang, Fuqi [email protected] Delft University of Technology, Köker, Derya Netherlands Alexion, United States [email protected] [email protected] Lokman, Banu Konur, Dincer University of Portsmouth, United Texas State University, United States Kingdom [email protected] [email protected] Korhonen, Pekka Lokman, Utku Aalto University, Finland University of Helsinki, United Kingdom [email protected] [email protected] Korkut, Kübra Macuada, Claudio Istanbul Technical University, Turkey Universidad de Santiago de Chile, Chile [email protected] [email protected] Köksalan, Murat Mantogiannis, Vasileios Middle East Technical University, Turkey Mott MacDonald, United Kingdom [email protected] [email protected] Krasny, Pawel Manyoma, Pablo EBRD, United Kingdom Universidad del Valle, Colombia [email protected] [email protected] Krátky, Tomáš Martins, Isabelle Centrum Hydraulického Vyzkumu spol. s SAGE/COPPE, Brazil r.o., Czech Republic [email protected] [email protected] Mateos, Alfonso Kucukaltan, Berk Universidad Politécnica de Madrid, Trakya University, Turkey Spain [email protected] [email protected] Kułakowski, Konrad Mebrek, Abdellah AGH UST, Poland Centre des techniques Spatiales, [email protected] Algeria [email protected]

189

Mechitov, Alexander Moslem, Sarbast University of Montevallo, United States Budapest University of Technology and [email protected] Economics, Hungary Medic, Nenad [email protected] University of Novi Sad, Serbia Mota, Caroline [email protected] Universidade Federal de Pernambuco, Medrano, F. Antonio Brazil Texas A&M University-Corpus Christi, [email protected] United States Moumni, Nourchen [email protected] University of Sfax, Tunisia Miclea, Adriana [email protected] Bucharest University of Economic Mousavi, Seyed Mohsen Studies, Romania University of Jyväskyla, Finland [email protected] [email protected] Miettinen, Kaisa Mu, Enrique University of Jyvaskyla, Finland Carlow University, United States [email protected] [email protected] Miroforidis, Janusz Mukhametzyanov, Irik Polish Academy of Sciences, Poland Ufa State Petroleum Technological [email protected] University, Russia Moraes, Fernanda [email protected] SAGE/COPPE, Brazil Nasrabadi, Nasim [email protected] University of Birjand, Iran Morales, Daniela [email protected] Universidad del Valle, Colombia Nishizaki, Ichiro [email protected] Hiroshima University, Japan Moralioglu, Begum [email protected] Istanbul Technical University, Turkey Nowak, Maciej [email protected] University of Economics in Katowice, Moshkovich, Elena Poland University of Montevallo, United States [email protected] [email protected] Onan, Kıvanc Doğuş University, Turkey [email protected]

190

Osheto, Shinji Özaydın, Özay Tohoku University, Japan Doğuş University, Turkey [email protected] [email protected] Oumaima, Khaled Özdemir, Mustafa Dassault Systèmes, France Eskisehir Technical University, Turkey [email protected] [email protected] Ouznadji, Said Özgen, İrem USTHB, Algeria Istanbul Technical University, Turkey [email protected] [email protected] Oyenuga, Olamilekan Özmez, Ali University of Lagos, Nigeria Doğuş Teknoloji, Turkey [email protected] [email protected] Ozok, Ahmet Fahri Özpeynirci, Selin Okan University, Turkey Izmir University of Economics, Turkey [email protected] [email protected] Ozpeynirci, Ozgur Öztayşi, Başar Izmir University of Economics, Turkey Istanbul Technical University, Turkey [email protected] [email protected] Ozturk, Onur Öztürk, Türker University of Ottawa, Canada Çimtaş, Turkey [email protected] [email protected] Ozturk, Onur Pelissari, Renata Çimtaş, Turkey UNICAMP, Brazil [email protected] [email protected] Önsel Ekici, Şule Pereyra-Rojas, Milagros Doğuş University, Turkey University of Pittsburgh, United States [email protected] [email protected] Özarslan, Ali Pessôa, Leonardo Antonio Middle East Technical University, Turkey CASNAV, Brazil [email protected] [email protected] Özateş Gürbüz, Melis Piercy, Craig Middle East Technical University, Turkey University of Georgia, United States [email protected] [email protected]

191

Pirlot, Marc Rezaei, Jafar University of Mons, Belgium Delft University of Technology, [email protected] Netherlands Podkopaev, Dmitry [email protected] Polish Academy of Sciences, Poland Rezaeinia, Nahid [email protected] Norwegian School of Economics, Pohl, Erik Norway Universität Duisburg-Essen, Germany [email protected] [email protected] Rodrigues, Márcio Polat, Seçkin Technical University of Liberec, Czech Istanbul Technical University, Turkey Republic [email protected] [email protected] Popović, Žarko Romero, Jorge University of Niš, Serbia Universidad Jorge Tadeo Lozano, [email protected] Colombia Popovici, Diana [email protected] University of Bucharest, Romania Rosenfeld, Jean [email protected] Université libre de Bruxelles, Belgium Porro, Olga [email protected] ESADE - URL, Spain Roumeissa, Kerbouı [email protected] USTHB, Algeria Qazi, Wajeeha [email protected] Middle East College, Oman Saaty, Rozann [email protected] Creative Decisions Foundation, United Raupp, Igor States Electric Energy Research Center - [email protected] CEPEL, Brazil Saaty, John [email protected] Decision Lens, United States Rauscher, Christian [email protected] Springer-Verlag GmbH, Germany Sahinkoc, Mert [email protected] Bogazici University, Turkey Reinhardt, Gilles [email protected] Université d'Ottawa, Canada Saini, Bhupinder [email protected] University of Jyvaskyla, Finland [email protected]

192

Salah, Mohamed Essalah Sınayiş, Nihat Can University of Sfax, Tunisia Istanbul Technical University, Turkey [email protected] [email protected] Samanlioglu, Funda Siraj, Sajid Kadir Has University, Turkey Leeds University Business School, [email protected] United Kingdom Sandholm, Tuomas [email protected] Carnegie Mellon, United States Skulimowski, Andrzej [email protected] AGH University of Science and Sarah, Kentache Technology, Poland USTHB, Algeria [email protected] [email protected] Sobrie, Olivier Savaş, Sezer University of Mons, Belgium Istanbul Technical University, Turkey [email protected] [email protected] Steuer, Ralph Sayın, Serpil University of Georgia, United States Koç University, Turkey [email protected] [email protected] Stewart, Theodor Schultes, Johanna University of Cape Town, South Africa University of Wuppertal, Germany [email protected] [email protected] Stiglmayr, Michael Schulze, Britta University of Wuppertal, Germany University of Wuppertal, Germany [email protected] [email protected] Swain, Anjan Kumar Sekizaki, Shinya Indian Institute of Management Hiroshima University, Japan Kozhikode, India [email protected] [email protected] Shavazipour, Babooshka Şahin Zorluoğlu, Özge University of Jyvaskyla, Finland Istanbul Technical University, Turkey [email protected] [email protected] Shayganmehr, Masoud Tezcaner Öztürk, Diclehan Tarbiat Modares University, Iran Hacettepe University, Turkey [email protected] [email protected]

193

Thalén, Björn Vargas, Luis Boeing, United States University of Pittsburgh, United States [email protected] [email protected] Tlili, Ali Vargas, Gwen Centrale Supelec, France United States [email protected] Walkowiak, Marek Topcu, Ilker Clemson University, United States Istanbul Technical University, Turkey [email protected] [email protected] Walkowiak, Olivia Trzaskalik, Tadeusz Clemson University, United States University of Economics in Katowice, [email protected] Poland Wallenius, Jyrki [email protected] Aalto University, Finland Tuncer Şakar, Ceren [email protected] Hacettepe University, Turkey Weidner, Petra [email protected] HAWK HHG, Germany Tureci, Hannan [email protected] Middle East Technical University, Turkey Wiecek, Margaret [email protected] Clemson University, United States Ucal Sari, Irem [email protected] Istanbul Technical University, Turkey Wu, Jei-Zheng [email protected] Soochow University, Taiwan Ulus, Firdevs [email protected] Bilkent University, Turkey Yanık, Seda [email protected] Istanbul Technical University, Turkey Unver, Berna [email protected] Sabancı University, Turkey Yanmaz, Ozgur [email protected] Istanbul Technical University, Turkey Uzturk, Deniz [email protected] Galatasaray University, Turkey Yılmaz, Beyza Özlem [email protected] Istanbul Medipol University, Turkey Ülengin, Füsun [email protected] Sabancı University, Turkey Yılmaz, Salim [email protected] Istanbul Technical University, Turkey [email protected]

194

Yunculer, Kutay Istanbul Technical University, Turkey [email protected] Zaabar, Imen École de Technologie Supérieure, Canada [email protected] Zhang, Jun Fudan University, China [email protected] Zhang, Yao Northeastern University, China [email protected] Zilci, Süleyman Çimtaş, Turkey [email protected]

195

Session Information

Session Session Time Room Identifier MON-0-O Opening Session Monday Mustafa 9:00-9:40 Kemal Amfisi MON-1-P Plenary Session 1 - Pınar Keskinocak Monday Mustafa 9:40-10:40 Kemal Amfisi MON-2-A Contributed Session: Advances in MCDM Theory - Monday Galata Tower Behavioral and Computational Issues 11:10-12:50 MON-2-B Invited Session: Theory and Applications of ANP Monday Maiden's 11:10-12:50 Tower MON-2-C Contributed Session: MCDM for Group Decisions Monday Dolmabahçe 11:10-12:50 Palace MON-2-D Special Session: AHP in Practice as ‘’Decision Monday Topkapı Conference’’ in Turkey - sponsored by ARAMA 11:10-12:50 Palace MON-2-E Invited Session: Data Science meets Multiple Monday Hagia Sophia Criteria Decision Making 11:10-12:50 MON-2-F Contributed Session: MCDM for Project Selection Monday Basilica 11:10-12:50 Cistern MON-3-A Contributed Session: Advances in MCDM Theory - Monday Galata Tower Applications in Diverse Industries 13:50-15:30 MON-3-B Contributed Session: Consistency Issues in AHP Monday Maiden's 13:50-15:30 Tower MON-3-C Invited Session: Building MCDM/A Models: Monday Dolmabahçe Practical and Methodological Issues 13:50-15:30 Palace MON-3-D Tutorial: Improve Your Decisions by Learning and Monday Topkapı Experiencing the ANP Best Practices- Part I 13:50-15:30 Palace MON-3-E Contributed Session: Recent Advances in Monday Hagia Sophia Multiobjective Optimization 13:50-15:30 MON-3-F Contributed Session: MCDM for Facility Location Monday Basilica and Logistics 13:50-15:30 Cistern MON-4-A Contributed Session: Advances in MCDM Theory Monday Galata Tower 16:00-17:40 MON-4-B Special Session: AHP/ANP Applications in Monday Maiden's Production and Manufacturing – sponsored by 16:00-17:40 Tower BORCELIK MON-4-C Contributed Session: Multiple Criteria Decision Monday Dolmabahçe Aiding 16:00-17:40 Palace MON-4-D Tutorial: Improve Your Decisions by Learning and Monday Topkapı Experiencing the ANP Best Practices- Part II 16:00-17:40 Palace MON-4-E Invited Session: Solution Methods for Various Monday Hagia Sophia Multiobjective Optimization Problems 16:00-17:40 MON-4-F Contributed Session: Multiobjective Monday Basilica Metaheuristics for Challenging Applications 16:00-17:40 Cistern

196

Session Session Time Room Identifier TUE-1-A Invited Session: MCDM/A for Assessment Issues Tuesday Galata Tower in Healthcare and LCA 9:00-10:40 TUE-1-B Contributed Session: AHP/ANP Applications in Tuesday Maiden's Academia and Education 9:00-10:40 Tower TUE-1-C Contributed Session: Preferences, Comparisons Tuesday Dolmabahçe and Uncertainties 9:00-10:40 Palace TUE-1-E Invited Session: Multiobjective Optimization Tuesday Hagia Sophia Applications 9:00-10:40 TUE-1-F Contributed Session: Understanding Consumers Tuesday Basilica and Markets 9:00-10:40 Cistern TUE-2-P Plenary Session 2 - Kathrin Klamroth Tuesday Mustafa 11:10-12:10 Kemal Amfisi TUE-3-A Contributed Session: Multiple Criteria Ranking or Tuesday Galata Tower Sorting 13:50-15:30 TUE-3-B Contributed Session: AHP/ANP Applications in Tuesday Maiden's Urban and Regional Development 13:50-15:30 Tower TUE-3-C Invited Session: MCDM/A Models with Flexible Tuesday Dolmabahçe and Interactive Preference Modeling 13:50-15:30 Palace TUE-3-D Contributed Session: Data Envelopment Analysis Tuesday Topkapı 13:50-15:30 Palace TUE-3-E Invited Session: Continuous Multiobjective Tuesday Hagia Sophia Optimization with Engineering Applications 13:50-15:30 TUE-3-F Contributed Session: Data, Applications and Tuesday Basilica MCDM 13:50-15:30 Cistern TUE-4-A Contributed Session: Multiple Criteria Ranking and Tuesday Galata Tower Sorting Methodology 16:00-17:40 TUE-4-B Contributed Session: Practical Applications of Tuesday Maiden's AHP/ANP 16:00-17:40 Tower TUE-4-C Contributed Session: Interactive Multiobjective Tuesday Dolmabahçe Optimization 16:00-17:40 Palace TUE-4-D Tutorial: MOLP vs. DEA - Relatives or Friends Tuesday Topkapı 16:00-17:40 Palace TUE-4-E Contributed Session: Advanced Optimization Tuesday Hagia Sophia Techniques and Multiple Objectives 16:00-17:40 TUE-4-F Invited Session: Multiobjective Route Planning Tuesday Basilica 16:00-17:40 Cistern WED-1-A Contributed Session: Outranking Methods Wednesday Galata Tower 9:00-10:40 WED-1-B Special Session: AHP/ANP Applications in Industry Wednesday Maiden's 4.0 - sponsored by ÇİMTAŞ 9:00-10:40 Tower WED-1-C Contributed Session: MCDM for Environmental Wednesday Dolmabahçe Decisions 9:00-10:40 Palace WED-1-D Contributed Session: Evolutionary Algorithms, Wednesday Topkapı Preferences and Applications 9:00-10:40 Palace WED-1-E Contributed Session: Portfolio Optimization Wednesday Hagia Sophia Models 9:00-10:40 WED-1-F Contributed Session: Practical Applications of Wednesday Basilica MCDM in an Interconnected World 9:00-10:40 Cistern

197

Session Session Time Room Identifier WED-2-P Plenary Session 3 - Tuomas Sandholm Wednesday Mustafa 11:10-12:10 Kemal Amfisi WED-3-P Doctoral Dissertation Awards Wednesday Mustafa 13:10-14:25 Kemal Amfisi THU-1-A Contributed Session: Multiple Criteria Ranking or Thursday Galata Tower Sorting 9:00-10:40 THU-1-B Contributed Session: AHP/ANP Applications in Thursday Maiden's Aviation 9:00-10:40 Tower THU-1-C Contributed Session: Contemporary Approaches in Thursday Dolmabahçe MCDM 9:00-10:40 Palace THU-1-E Contributed Session: Bringing Multiobjective Thursday Hagia Sophia Optimization to Applications 9:00-10:40 THU-1-F Contributed Session: Practical Applications of Thursday Basilica MCDM 9:00-10:40 Cistern THU-2-P Plenary Session 4 - Luis Vargas Thursday Mustafa 11:10-12:10 Kemal Amfisi THU-3-U Business Meeting Thursday Mustafa 13:10-14:10 Kemal Amfisi THU-4-A Invited Session: Hybrid Approaches in MCDM Thursday Galata Tower 14:20-16:00 THU-4-B Contributed Session: Business Applications of Thursday Maiden's AHP/ANP 14:20-16:00 Tower THU-4-C Invited Session: MCDM in Solving Urban Problems Thursday Dolmabahçe using Big Data 14:20-16:00 Palace THU-4-E Contributed Session: Practical Approaches to Thursday Hagia Sophia Challenging Problems in Multiobjective 14:20-16:00 Optimization THU-4-F Contributed Session: MCDM for Strategic, Tactical Thursday Basilica and Operational Decisions 14:20-16:00 Cistern FRI-1-M Society Award Talks Friday 9:30- Galata Tower 12:00 FRI-2-A Invited Session: New Approaches in MCDM Friday 13:00- Galata Tower 14:40 FRI-2-B Contributed Session: Fuzzy sets and approaches Friday 13:00- Maiden's 14:40 Tower FRI-2-C Contributed Session: Advances in MCDA Friday 13:00- Dolmabahçe 14:40 Palace FRI-2-E Contributed Session: Evolutionary Algorithms and Friday 13:00- Hagia Sophia Multiobjective Optimization 14:40 FRI-2-F Contributed Session: Making Decisions with Friday 13:00- Basilica Environmental Concerns 14:40 Cistern FRI-3-L Closing Session Friday 14:40- Galata Tower 15:10

198