Forecasting Techniques in Construction Industry: Earned Value Indicators and Performance Models

Scientiﬁ c Review – Engineering and Environmental Sciences (2020), 29 (2), 234–243 Sci. Rev. Eng. Env. Sci. (2020), 29 (2) Przegląd Naukowy – Inżynieria i Kształtowanie Środowiska (2020), 29 (2), 234–243 Prz. Nauk. Inż. Kszt. Środ. (2020), 29 (2) http://iks.pn.sggw.pl DOI 10.22630/PNIKS.2020.29.2.20 Firas Kh. JABER1, Nidal A. JASIM2, Faiq M.S. Al-ZWAINY3 1 Middle Technical University, Electrical Engineering Technical College 2 University of Diyala, College of Engineering 3 Al-Nahrain University, College of Engineering Forecasting techniques in construction industry: earned value indicators and performance models

Key words: Machine Learning Regression other industries by its many risks, and its Techniques (MLRT), earned value indexes, projects always suffer from the problems SPI, CPI, and TCPI of delay in implementation and increase in cost in most countries of the world. Among the most important characteris- Introduction tics of the construction industry (Myers, 2005): The construction industry is an im- 1. The nature of its product is unique, portant industry for any government each project differs from the other, due to its direct association with the im- and the temporary of each project is plementation of the goals and policies limited in duration and location, with of the government in various ﬁ elds of the completion of the project, the concern to the citizen in terms of educa- equipment and labor will be trans- tion, health, housing and other facilities ferred to another project in another and services. The construction industry place. is also one of the broad and important 2. The nature of work within a single sectors of any country’s economy, and it project is fragmented, as several dif- is one of the main engines that govern- ferent parties separate and separate ments resort to move the economy and to complete the project. create jobs and reduce unemployment. 3. The industry’s heavy dependence on And the construction industry has unique manpower and its lack of research characteristics, which made it an industry and development in methods and that is distinguished from the rest of the methods of implementation.

234 F.Kh. Jaber, N.A. Jasim, F.M.S. Al-Zwainy 4.The criterion of competition for bid- cost status. Earned value management ding in most cases is the lower price, system in effect integrates time manage- which reduces the profi t margin and ment and cost management which are es- thus reduces the desire of the compa- sential elements of tall buildings projects nies executing in research, develop- management (Al-Zwainy, Mohammed & ment and creativity. Mohsen, 2015). The increasing demand for tall Review of the literature related to buildings represents a huge opportunity earned value management system with to reconstruct the construction industry Forecasting techniques such as Ma- and create value through joint action. chine Learning Regression Techniques According to research prepared by the (MLRT) was in top management jour- McKinsey Global Institute, the industry nals such as International Journal of can boost the productivity of its work- Project Management. Where, various ers by up to 60% if changes are made researchers have used Machine Learn- in seven key areas: regulations, design ing Regression Techniques (MLRT) as operations, contracts, supply and supply a tool for prediction and optimization chain management, and on-site imple- in different project management ar- mentation (Granskog, Guttman & Sjö- eas are also reviewed, including earned din, 2016). value management. The great majority Tall buildings since the 1990s and of project management applications of during the fi rst decade of the 21st cen- Multiple Linear Regression are based on tury are considered civilization and ur- the inter algorithm. The previous studies banization facilities, not only that, but are among the most important scientifi c tall buildings have become symbols of foundations on which the current study countries. Tall buildings are defi ned as is based. The researcher begins with re- those buildings whose height is not less search, examination and analysis in the than 40 m or 13 fl oors. In most cases, the previous studies; this is because the re- management of tall projects is not a sim- search processes are cumulative process- ple matter, as success in such projects es that depend on the previous ones. In requires good coordination as well as the next paragraph, the previous studies a comprehensive integrated operating on the earned value management will be system for the project, so that all project addressed and techniques used for fore- participants can understand their roles casting purposes. and agree on key performance indicators During the last few years or so, the (Al-Zwainy & Edan, 2017). use of EVM has increased in many con- Earned value management system struction engineering projects and has (EVMS) is a useful management tech- demonstrated some degree of success. nique available for tall buildings projects The following literature review reveals managers to monitor and control projects; that EVM have been used successfully in a technique that combines the work modelling of the earned value manage- scope, schedule and the cost elements ment. Pajares and Lopez-Paredes (2011) of a project and facilitates the integrated proposed two new metrics that combine reporting of a project’s progress and the earned value management and project

Forecasting techniques in construction industry: earned value indicators... 235 risk management for construction project neurial orientation and entrepreneurial controlling and monitoring. Elshaer attitude of the individuals. (2013) studied the impact of the activi- The researchers did not found any ties sensitivity information on the fore- study that is exactly the same as the cur- casting accuracy of the Earned Schedule rent study while searching in interna- Method (ESM). Czemplik (2014) applied tional libraries and discreet periodicals schedule forecast Indicator (SFI) to sup- such as Scopus, Springer, Taylor France port site managerial decision concerning and others, as this current study is almost variation orders. Khamooshi and Golaf- unique of its kind that deals with predict- shani (2014) provided a new approach to ing the value indicators gained using decouple the schedule and cost dimen- Machine Learning Regression Tech- sions in EVM by adding earned duration niques (MLRT) in tall building projects, while maintaining the unique interaction despite the existence of multiple studies of the three major project management dealing with a topic prediction in con- elements of scope, cost, and time similar struction projects. Based on the above to EVM. Chen, Chen and Lin (2016) im- discussion the contribution of this study proved the predictive power of Planned are as follows. Value (PV) before executing construction Firstly, in this study, the main tar- project. Jaber, Hachem and Al-Zwainy get is on using associate Forecasting (2019) applied the value management techniques for the aim of earned value methodology acquired in the waste wa- prediction, meant as a decision-support ter treatment plant projects to measure aid to project manager, contractors and performance by predicting performance planners. The contribution of this study indicators such as CPI, SPI and TCPI. is to spot the most effective activity in The following literature review re- terms of model structure and parameters veals that Machine Learning Regression on an earned value prediction. Applied Techniques (MLRT) have been used suc- interest of this study may well be use cessfully in construction project manage- machine intelligence techniques such ment (Al-Zwainy et al., 2015) developing as Machine Learning Regression Tech- EAC model for bridges projects using niques (MLRT) as an acceptable tool for Multifactor Linear Regression Technique earned value estimating, where, Project (MLRT). Bilal and Oyedele (2020) pre- managers in Iraq; typically have to be sented guidelines for Applied Machine compelled to estimate the earned value Learning (AML) in the construction in- of construction projects at construction dustry from training to operationalizing stage quickly and around to supply fund- models, which are drawn from our ex- ing or to get the adoption of the budget perience of working with construction from decision-makers. Therefore, it’s folks to deliver Construction Simulation necessary to search out the simplest way Tool (CST). Sabahi and Parast (2020) to understand the earned value of the used predictive analytics by proposing a construction projects in a very short time machine learning approach to predict in- with acceptable accuracy. dividuals’ project performance based on In summary, the application EVM in measures of several aspects of entrepre- tall building projects would be speciﬁ -

236 F.Kh. Jaber, N.A. Jasim, F.M.S. Al-Zwainy cally helpful for a state like Iraq which SPSS is short for Statistical Package suffers from shortage of a high number for the Social Sciences, and it’s utilized of severe tall buildings. The main ob- by different sorts of analysts for com- jective of this study is to develop Ma- plex measurable information investiga- chine Learning Regression Techniques tion. The SPSS programming bundle (MLRT) to predict the earned value in- was made for the administration and fac- dicators at early stage of tall buildings tual examination information in different projects in Iraq to reduce the percentage sciences. Long produced by SPSS Inc., error of estimation. This will facilitate it was acquired by IBM in 2009. The the project practitioners to the further use current versions are named IBM SPSS of EVA for the schedule and cost control Statistics. of their construction projects. To achieve this, there is a need to identify the factors that affect the performance of residential Identifi cation ANN Models buildings projects that can be available variables at early stages. Therefore, the specialists researchers in this study are endeavor- The Bismayah City Project is the fi rst ing improvement and assessment earned of its kind and the largest in the history value models through the below stages: of Iraq in terms of size and level of ser- 1. Choose the appropriate statistics vices and infrastructure provided in it, software. and Hanwha Engineering and Construc- 2. Identifi cation of MLRT variables that tion Company is proud to be the fi rst have an effect on the earned value in- company to implement such a strategic dex in tall buildings projects. project in Iraq. The new city of Basmaya 3. Building a numerical model to fore- is an integrated residential complex lo- see of the earned value indexes (SPI, cated 10 km southeast of the capital, SPI and TCPI). Baghdad, Iraq, on the international road 4. Check the confi rmation and approval No 6 linking the capital Baghdad and of the scientifi c models. Wasit Governorate, and on an area of 1,830 ha, and consists of 100,000 hous- Choose the appropriate statistics ing units, and also includes a network In- software tegrated infrastructure of water and elec- tricity, as well as schools and recreational Several applications that support the and commercial complexes. The cost of establishment of statistics analysis like constructing the complex amounts to Microsoft Excel, Statistica, MINITAB, $7.75 billion, the project is implemented and MATLAB, but this study was se- by the Korean Hanwha Engineering and lected SPSS Program, where SPSS is Construction Company, which works in the premier statistics analysis environ- the fi eld of construction and construction ment. The SPSS range provides an easy- around the world, with a duration of im- -to-use, visual, object-oriented approach plementation of up to seven years. The to problem solving using intelligent new Basmaya city represents a new mod- technologies. ern lifestyle that may be characterized by

Forecasting techniques in construction industry: earned value indicators... 237 a high level of services and an ideal envi- dex (CPI), Schedule Performance Index ronment suitable for the development of (SPI) and To Complete Cost Perform- new generations. The aerial photos show ance Indicator (TCPI) are deﬁ ned as the the level of progress of work and com- dependent variables. There are many pletion of other residential and service variables as an independent variable such buildings, as shown in Figure 1. as: Budget at Completion (BAC), Actu- al Cost (AC), Actual Percentage (A%), Earned Value (EV), Planning Percentage (P%), and Planning Value (PV).

Building Machine Learning Regression (MLR) Models

The factors that were recognized at the information distinguishing proof stage FIGURE 1. The new city of Basmaya in Iraq were utilized to build up the Machine Learning Regression Techniques (MLRT). Machine Learning Regression Tech- Three numerical models were created in niques (MLRT) requires a lot of data. this examination by utilizing Machine Therefore, a lot of historical tall build- Learning Regression Techniques (MLRT). ings projects were collected which were It was utilized on the grounds that it is prob- done between 2012 and 2018 as shown ably the most kind that broadly utilized in in Table 1. There are two types of vari- this fi eld of study. What’s more, the ven- ables that affected on the earned value in ture qualities or parameters in a numeri- tall buildings projects in Republic of Iraq cal model were utilized to foresee earned which are dependent variables and inde- value indexes. The Statistical Product and pendent variables. Cost Performance In- Solutions Services (SPSS) programming forms 26 are utilized as a tool to build up TABLE 1. Summary of historical data for Bis- the three models, as following: mayah City Project 1. Schedule Performance Index (SPI). 2. Cost Performance Index (CPI). Parameter Max. Min. Range AVG 3. To Complete Cost Performance Indi- BAC 950 800 150 865 cator (TCPI). AC 600 200 400 375 A% 60 25 35 39 EV 510 237.5 272.5 339.8 First: Schedule Performance P% 60 25 35 42 Index Model PV 510 237.5 272.5 358 SPI 1.11 0.75 0.36 0.95 Table 2 expounds multiple linear re- CPI 1.18 0.80 0.38 0.93 gression estimates, such as coeffi cients of standardized, coeffi cients of unstand- TCPI 0.80 1.19 0.80 0.39 ardized, standardized errors, t-test and

238 F.Kh. Jaber, N.A. Jasim, F.M.S. Al-Zwainy TABLE 2. Coeffi cients of SPI-Model (dependent variable: Y1) Standardized Unstandardized coeffi cients Model coeffi cients t sig. BSEß Constant –14.393 8.593 – –1.675 0.193 X1 0.018 0.010 3.253 1.784 0.172 X2 0.007 0.005 2.684 1.255 0.298 X3 –0.076 52.969 –0.026 –0.001 0.999 X4 –0.013 0.066 –3.922 –0.197 0.857 X5 33.633 50.328 10.067 0.668 0.552 X6 –0.034 0.057 –8.747 –0.590 0.597 testing the signifi cant of every autono- Statistics of regression for model CPI mous factors or independent variable. are set in Table 3, which can be formed The regression statistics of Schedule as Equation (2): Performance Index Model (SPI-Model) are set in Table 1, which can be formed Y2 = CPI = –12.405 + (0.015X1) + as Equation (1): + (0.002X2) – (22.344X3) + (0.016X4) + + (50.059X5) – (0.05X6) (2) Y1 = SPI = –14.393 + (0.018X1) + + (0.007X2) – (0.076X3) – (0.013X4) + + (33.633X5) – (0.034X6) (1) Third: Complete Cost Performance Indicator Second: Cost Performance Index Model The regression statistics of To Com- plete Cost Performance Indicator (TCPI- The regression statistics of Cost Per- -Model) are set in Table 4. formance Index Model (CPI-Model) are set in Table 3. TABLE 3. Coeffi cients of CPI-Model (dependent variable: Y2) Standardized Unstandardized coeffi cients Model coeffi cients t sig. BSEß Constant –12.405 7.813 – –1.588 0.211 X1 0.015 0.009 2.609 1.641 0.199 X2 0.002 0.005 0.881 0.472 0.669 X3 –22.344 48.16 –7.317 –0.464 0.674 X4 0.016 0.060 4.731 0.272 0.803 X5 50.059 45.75 14.372 1.094 0.354 X6 –0.050 0.052 –12.444 –0.963 0.407

Forecasting techniques in construction industry: earned value indicators... 239 TABLE 4. Coeffi cients of CPI-Model (dependent variable: Y3) Standardized Unstandardized coeffi cients Model coeffi cients t sig. BSEß Constant –1.436 8.477 – –0.169 0.876 X1 0.001 0.010 0.164 0.092 0.932 X2 –0.002 0.005 –0.789 –0.378 0.731 X3 –31.521 52.258 –10.647 –0.603 0.589 X4 0.036 0.065 10.635 0.547 0.623 X5 31.441 49.653 9.311 0.633 0.572 X6 –0.030 0.057 –7.538 –0.521 0.638

Statistics of regression for model the estimation of standard error. This TCPI are set in Table 4, which can be statistical analysis was bring about for formed as Equation (3): the MLR models (SPI, CPI, and TCPI) between output (earned value indexes) Y3 = TCPI = –1.436 + (0.001X1) – and input (BAC, Budget at Completion – (0.002X2) – (31.521X3) + (0.036X4) + – X1; AC, Actual Cost – X2; A%, Actual + (31.441X5) – (0.03X6) (3) Percentage – X3; EV, Earned Value – X4; P%, Planning Percentage – X5; PV, Plan- ning Value – X6). On the other hand, the Check the Confi rmation and R-values for SPI, CPI, and TCPI models Approval of the Scientifi c Models are equal to 85.5%, 89.2%, and 86.3% respectively, which is viewed as extremely Below is the model’s summary, aloft correlation. In addition, the deter- which has some important statistics. It mination coeffi cient shows the extent provides us correlation coeffi cient (R) of the variety in input variable from an with determination coeffi cient (R2) and output variables, and that models have

1 0.892 0.863 0.855 0.9 0.796 0.744 0.732 0.8 0.7

SPI 0.6

CPI 0.5 TCPI 0.4 0.283 0.279 0.257 0.3

0.2

0.1 SE R2 R FIGURE 1. Summary statistics of models

240 F.Kh. Jaber, N.A. Jasim, F.M.S. Al-Zwainy R2 values equal to 73.2%, 79.6%, and 74.4% respectively. From Figure 1, it can be seen that model (SPI), model (CPI), model (TCPI) verifi cation have pretty performance in order that offered high correlation of coeffi cient (R) 85.5%, 89.2% and 86.3%, respectively. Figures 2–4 illustrate the capability of MLR models (SPI-model, CPI-model, TCPI-model), where, the coeffi cient of determination (R2) is 73.2%,

FIGURE 4. Relationship between observed TCPI and expected TCPI

79.6% and 74.4%, therefore, it can be concluded that these models shows an excellent agreement with the actual measurements.

Conclusions

FIGURE 2. Relationship between observed SPI In this study, a depiction of models and expected SPI factors, information assortment strat- egy and factual determination of verifi able information has been clarifi ed. Bismayah New City was selected as a case study, six variables were adopts as independent variables (BAC, Budget at Completion – X1; AC, Actual Cost – X2; A%, Actual Percentage – X3; EV, Earned Value – X4; P%, Planning Percentage – X5; PV, Planning Value – X6), and three variables were (Cost Performance Index – CPI; Schedule Performance In- dex – SPI; To Complete Cost Perform- ance Indicator – TCPI) are defi ned as the dependent variables. In addition, Ma- chine Learning Regression Techniques FIGURE 3. Relationship between observed CPI (MLRT) has been used to build predic- and expected CPI tion model for earned value indexes by

Forecasting techniques in construction industry: earned value indicators... 241 using SPSS-26 statistically software. It putations. Nidal A. Jasim confi rmed was found that the MLRT showed good the logical techniques and regulated the results of estimation in terms of correla- discoveries of this work. All researchers tion coeffi cient (R) generated by MLR talked about the outcomes and writting models for SPI and CPI and TCPI where to the last composition of manuscript. the R were 85.5%, 89.2%, and 86.3% respectively. At long last, a result tends to be presumed that these models show References a brilliant concurrence with the genuine estimations. Al-Zwainy, F.M.S. & Edan, I.A. (2017). Infor- mation technology in construction project The availability of data management. Amman, Jordan: Darghaidaa for Publishing [translated from Arabic]. The study discoveries which were Al-Zwainy, F.M.S., Mohammed, I. & Mohsen, bolstered by the information, are realistic D.S. (2015). Earned value management in by the relating’s essayist can be obtained construction project. Saarbrűcken: LAP on demands. LAMBERT Academic Publishing. Bilal, M. & Oyedele, L.O. (2020). Guidelines for applied machine learning in construction Confl icts of interest industry – a case of profi t margins estima- The authors asserted that they hadn’t tion. Advanced Engineering Informatics, 43, interest clashes and confl icts. 101013. DOI 10.1016/j.aei.2019.101013 Chen, H.L., Chen, W.T. & Lin, Y.L. (2016). Earned value project management: improv- Acknowledgements ing the predictive power of planned value. Authors would like to thank those International Journal of Project Manage- ment, 34(1), 22-29. respected persons who have helped and Czemplik, A. (2014). Application of earned value guided us to achieve this paper; they re- method to progress control of construction ally deserve all respect and gratitude. projects, Procedia Engineering, 91(1), Then, we’d like to thank all the staff of 424-428. the Al-Nahrain University, Middle Tech- Elshaer, R. (2013). Impact of sensitivity informa- nical University, Baghdad, and the Uni- tion on the prediction of project’s duration using earned schedule method. International versity of Diyala in Iraq to help us to get Journal of Project Management, 31(4), guideline for our paper by providing us 579-588. lots of consultations. As well as, we also Granskog, A., Guttman, B. & Sjödin, E. (2016). want to thank and give our gratitude to Toward a customer-centric construction- everyone who has guided or advised us equipment industry. McKinsey & Company. in submitting or writing this paper. Retrieved from: https://www.mckinsey.com/ industries/automotive-and-assembly/our-in- sights/toward-a-customer-centric-construc- Author contributions tion-equipment-industry Firas Kh. Jaber conceived of the pre- Jaber, F.K., Hachem, S.W. & Al-Zwainy, F.M. sented idea and contributed to sample (2019). Calculating the indexes of earned value for assessment the performance of preparation and wrote the manuscript. waste water treatment plant. ARPN Journal Faiq M.S. Al-Zwainy developed the of Engineering and Applied Sciences, 14(4), MLR models and performed the com- 792-802.

242 F.Kh. Jaber, N.A. Jasim, F.M.S. Al-Zwainy Khamooshi, H. & Golafshani, H. (2014), EDM: see Schedule Performance Index (SPI) as Earned Duration Management, a new ap- ﬁ rst model, Cost Performance Index (CPI) proach to schedule performance management as a second model, and the third model is and measurement. International Journal of To Complete Cost Performance Indicator Project Management, 32(6), 1019-1041. (TCPI) in Bismayah New City was chosen Myers, D. (2005). Construction economics: a new as a case study. The methodology is mainly approach. London: Spon Press. impacted by the deciding various compo- Pajares, J. & Lopez-Paredes, A. (2011). An nents (variables) which impact on the earned extension of the EVM analysis for project value analysis, six free factors (X1: BAC, monitoring: The Cost Control Index and the Budget at Completion; X2: AC, Actual Cost; Schedule Control Index. International Jour- X3: A%, Actual Percentage; X4: EV, Earned nal of Project Management, 29(5), 615-621. Sabahi, S. & Parast, M.M. (2020). The impact of Value; X5: P%, Planning Percentage, and entrepreneurship orientation on project per- X6: PV, Planning Value) were self-assertive- formance: a machine learning approach. In- ly assigned and agreeably depicted for per ternational Journal of Production Economics, tall buildings projects. It was found that the 107621. DOI 10.1016/j.ijpe.2020.107621 MLRT showed good results of estimation in terms of correlation coefﬁ cient (R) generated by MLR models for SPI and CPI and TCPI where the R were 85.5%, 89.2%, and 86.3% Summary respectively. At long last, a result tends to be presumed that these models show a brilliant Forecasting techniques in construc- concurrence with the genuine estimations. tion industry: earned value indicators and performance models. Machine Learning Regression Techniques (MLRT) as a shrewd Authors’ address: method can be utilized in this study being Faiq M. S. Al-Zwainy exceptionally fruitful in demonstrating non- (https://orcid.org/0000-0002-9948-6594) -linear and the interrelationships among Al-Nahrain University them in problems of construction projects College of Engineering such as the earned value indexes for tall Department of Civil Engineering buildings projects in Republic of Iraq. Three 10072 Al-Jadryia, Baghdad, Iraq forecasting models were developed to fore- e-mail: [email protected]

Forecasting techniques in construction industry: earned value indicators... 243