Vol. 6, No. 1, March 2013

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Volume 6. No. 1. March, 2013

March, 2013 www.ajocict.net

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Volume 6. No. 1. March, 2013

March, 2013 www.ajocict.net

A Journal of the Institute of Electrical & Electronics Engineers (IEEE) Computer Chapter Nigeria Section

ISSN- 2006-1781

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Published Papers

1-9 Forecasting Gas Flaring Pollution Concentration with Neural Networks O.S. Asaolu Department of Systems Engineering, University of Lagos, Akoka, Lagos, Nigeria

7-20 Empirical Evaluation of Customers’ Use of Electronic Banking Systems in Nigeria Onyedimekwu Okechi and Oruan Memoye Kepeghom AfriHUB ICT Solution for Africa, Federal college of Education, P.M.B 11, Omoku, Rivers state.

21-32 Design and Implementation of a Microcontroller Based Automatic Gate O. Shoewu & Segun O. Olatinwo Department of Electronic and Computer Engineering, Lagos State University, Epe Campus, Nigeria.

33-42 Understanding the Potential of Data Mining in Botswana George Anderson, Audrey N. Masizana-Katongo, and Dimane Mpoeleng Department of Computer Science, University of Botswana, Gaborone, Botswana

43-48 A Framework For Knowledge Based Ontology Model In African Traditional Medicine Omotosho, L.O.; **Odejobi, O.A. and +Akanbi C.O. Department of Computer Science, Afe Babalola University, Ado-Ekiti, Nigeria.

49-58 Design and Implementation of an Enhanced Power Billing System for Electricity Consumers in Nigeria Adegboye Adegboyega, Ayeni A. Gabriel, Alawode .J. Ademola & Azeta .I. Victor Department of Computer Science, Achievers University, Owo, Nigeria. E-Mail: [email protected]

59-68 Better Quality Of Service Management With Fuzzy Logic In Mobile Adhoc Network Onifade O.F.W.,Ojesanmi O.A. and Oyebisi T.O. Department of Computer Science, University of Ibadan, Nigeria.

69-78 Evaluating Usability Factors In Different Authentication Methods Using Artificial Neural Network Mapayi, T., Olaniyan, O.M, Isamotu, N.O., Raji, S.O. & Olaifa, M. Department of Computer Science University of KwaZulu-Natal Durban, South Africa

79-86 Modeling Confidentiality Archetype and Data Mining in Cloud Computing Alawode A. Olaide Imo State University, Owerri, Nigeria

87-94 Comparison of Speech and DTMF for VoiceXML-Based Expert System: Implications for Developers Oyelami Olufemi Moses, Uwadia Charles Onuwa, Akinwale Adio Taofeek & Akinyemi Ibidapo Olawole Department of Computer and Information Sciences, Covenant University, Ota, Nigeria

95-108 Development of an Automated Parking Lot Management System Segun O. Olatinwo and O. Shoewu Department of Computer Science and Engineering, Ladoke Akintola University of Technology, Ogbomoso, Nigeria.

109-116 A Survey of Techniques For Answering Top-k Queries Neethu C V Dept. of Computer Science & Engineering, SCT College of Engineering, Trivandrum,India

iii

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

117-126 Performance Evaluation of AN-VE: An Error Detection and Correction Code Egwali Annie O. & Akwukwuma V. V. N. Department of Computer Science, Faculty of Physical Sciences. University of Benin. P.M.B. 1154. Benin City. Nigeria.

127-140 Evolutionary Continuous Genetic Algorithm for Clinical Decision Support System Prem Pal Singh Tomar & Ranjit Singh1, Faculty of Engineering, Dayalbagh Educational Institute, Agra, India

141-146 Mass Connectivity based Digital Nervous System for Nigeria. Monica N. Agu Department of Computer Science, University of Nigeria, Nsukka, Nigeria.

147-164 Performance Evaluation of Neural Network MLP and ANFIS models for Weather Forecasting Studies Oyediran O. F. and Adeyemo A. B. Computer Science Department, University of Ibadan

165-172 Web Designers Guide on Development Technologies: An Evaluation Approach F.E. Ekpenyonga * and D. T. Chinyio Department of Mathematics and Computer Science, Nigerian Defence Academy Kaduna.

173-178 Adaptive estimation of low frequency noise in Impedance Cardiography Ms. Madhavi Mallam1, Dr.A.GuruvaReddy1 Department of ECE, DVR & Dr.HS MIC College of Technology, Kanchikacharla.

179-184 Stochastic Optimization Techniques as Effective Tools to Load Forecasting and Sxcheduling Using Energy Resources (DERs) C.G. Monyei Department of Electrical Engineering, University of Ibadan, Ibadan, Nigeria

185-190 Introducing Spatial Qualification Problem and Its Qualitative Model P.C. Bassey & Akinkunmi B.O. Department of Computer Science, University of Ibadan, Ibadan, Nigeria.

191-202 An Empirical Investigation of the Level of Adoption of Mobile Payment in Nigeria Adebiyi, A.A., Alabi, E. Ayo, C.K. and Adebiyi, M.O. Department of Computer Science, Covenant University, Ota, Nigeria

203-210 Internet Banking Authentication Methods in Nigerian Commercial Banks O.B. Lawal, A. Ibitola & O.B. Longe Olabisi Onabanjo University Consult, Ibadan, Nigeria

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Editorial Board

Editor-in-Chief

Prof. Dele Oluwade Senior Member (IEEE) & Chair IEEE Nigeria – Computer Chapter. College of Information & Communication Technology Salem University, Lokoja, Nigeria

Editorial Advisory Board

Prof. Gloria Chukwudebe - Senior Member & Chairman IEEE Nigeria Section

Engr. Tunde Salihu – Senior Member & Former Chairman IEEE Nigeria Section

Prof. Adenike Osofisan - University of Ibadan, Nigeria

Prof. Amos David – Universite Nancy2, France

Prof. Clement K. Dzidonu – President Accra Institute of Technology, Ghana

Prof. Adebayo Adeyemi – Vice Chancellor, Bells University, Nigeria

Prof. S.C. Chiemeke – University of Benin, Nigeria

Prof. Akaro Ibrahim Mainoma – DVC (Admin) Nasarawa State University, Nigeria

Dr. Richard Boateng – University of Ghana, Ghana.

Prof. Lynette Kvassny – Pennsylvania State University, USA

Prof. C.K. Ayo – Covenant University, Nigeria

Dr. Williams Obiozor – Bloomsburg University of Pennsylvania, USA

Prof Enoh Tangjong – University of Beau, Cameroon

Prof. Sulayman Sowe, United Nations University Institute of Advanced Studies, Japan

Dr. John Effah, University of Ghana Business School, Ghana

Mr. Colin Thakur - Durban University of Technology, South Africa

Mr. Adegoke, M.A. – Bells University of Technology, Ota, Nigeria

Managing/Production Editor

Dr. Longe Olumide PhD Department of Computer Science University of Ibadan, Ibadan, Nigeria

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Foreward

The African Journal of Computing & ICT remains at the nexus of providing a platform for contributions to discourses, developments, growth and implementation of Computing and ICT initiatives by providing an avenue for scholars from the developing countries and other nations across the world to contribute to the solution paradigm through timely dissemination of research findings as well as new insights into how to identify and mitigate possible unintended consequences of ICTs. Published papers presented in this voume provide distinctive perspective on practical issues, opportunities and dimensions to the possibilities that ICTs offer the African Society and humanity at large. Of note are the increasing multi-disciplinary flavours now being demonstrated by authors collaborating to publish papers that reflect the beauty of synergistic academic and purpose-driven research. Obviously, these developments will drive growth and development in ICTs in Africa.

This issue of the African Journal of Computing & ICTs contains journal articles with a variety of perspective on theoretical and practical research conducted by well-grounded scholars within the sphere of information technology and allied fields across the globe. While welcoming you to peruse this volume of the African Journal of Computing and ICTs, we encourage you to submit your manuscript for consideration in future issues of the Journal

Have a fruitful reading

Thank you

Longe Olumide Babatope PhD Managing Editor Afr J Comp & ICTs

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Algorithms for Verifying Variants of Boolean Algebra Equations Forecasting Gas Flaring Pollution Concentration Using Neural Networks

O.S. Asaolu Department of Systems Engineering University of Lagos Akoka, Lagos, Nigeria [email protected]

ABSTRACT Gas flaring emissions constitute a threat to both terrestrial and aquatic life. This work establishes the viability of Neural Networks for forecasting concentrate volumes and is situated within Missing Data and Quality problems. A feed-forward Multi-Layer Perceptron (MLP) with 2 hidden layers (consisting of 200 nodes each) and a multiple output layer was developed using Matlab software. The six input nodes models Stack height, Temperature ratio, Flame length, Gas flow rate, Hydrogen Sulphide in solution gas and Combustion efficiency. The eight output nodes give the percentage volume ratio of emission concentrates. Data was collated from Nigerian Oil & Gas firms and the Federal Environmental Protection Agency. The neural network inputs and targets of the 264 sample training sets were scaled by normalizing the mean (to 0) and standard deviation (to 1). The outputs; Benzene, Toluene, Carbon, C3HC, C4HC, Styrene, Acetylene and Xylene were found to have a volume ratio of 6%, 13%, 12%, 13%, 18%, 13%, 12% and 13% respectively. The simulated network outputs were plotted against the targets as open circles. The slopes of linear fit indicated fair correlation between targets and outputs. This ANN prediction could aid the formulation of polices to check gas flaring such as permissible levels of pollutant constituents over time so as to check hazardous exposures and environmental degradation.

Keywords- Neural Networks, ANN, Gas flare, Flaring, Forecasting African Journal of Computing & ICT Reference Format: O.S. Asaolu (2013). Forecasting Gas Flaring Pollution Concentration Using Neural Networks. Afr J. of Comp & ICTs. Vol 6, No. 1. pp 1- 6

1. INTRODUCTION Table 1.0: Top 20 Flaring Countries A gas flare or flare stack is an elevated chimney for burning off flammable gas and pressurerized liquids released from plant process. In Nigeria, flaring of gas mixed up with the crude oil began right at the beginning in 1958, and so did recognition of its unacceptability. Production flares have contributed a lot of greenhouse gases containing toxins that affect the health and livelihood of the local communities as well as the ecosystem. During combustion, several intermediate products, mostly hydrocarbons are formed, and eventually released into the atmosphere, ultimately contributing to the global warming menace.

In a country with a high activity of petroleum exploration and production, gas flaring though inevitable should be well managed and minimized. Table 1.0 show that Nigeria is one of the world’s leading gas flaring nations. [1] reported that the Nigerian legislature keeps revising deadlines to end flaring and the sanctions on companies that fail to comply. Nigeria is ranked as holding the world’s seventh largest gas reserves with proven reserves at about 187 trillion cubic feet but flares about 30% of its current daily gas production of 4 billion cubic feet, losing $2.5 billion every year in revenue. Source: World Bank GGFR Report [3]

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The people of Niger Delta region in Nigeria turned to militancy in part because of inadequate government effort to redress environmental damage caused by oil exploration activities. Their livelihoods based on agriculture and fishing are perpetually under threat from oil spills, gas flaring, acid rain and other forms of despoliation, as noted in [2]. Launched at the World Summit on Sustainable Development in August 2002, the Global Gas Flaring Reduction public-private partnership (GGFR) brings around the table representatives of governments of oil-producing countries, state-owned companies and major international oil companies so that together they can overcome the barriers to reducing gas flaring by sharing global best practices and implementing country specific programs.

2. RELATED WORK

Recent studies [4] involving laboratory evaluation of air, rain and water samples in the Niger Delta region established that flow-station emissions significantly impacted the environment and thus require mitigation measures to avert inherent biomagnifications with time. Figure 1: A three-layer neural network A system of forecasting pollution concentration would help in formulation of polices to control and improve ways checking the menace. Many researchers have ANN has been widely used all over the world to predict worked in the area artificial neural networks and time series pertaining to various kinds of pollution [5, 6]. forecasting of air pollutants. Indeed [7] compared the performance artificial neural network over multiple regression models in predicting Artificial Neural Networks (ANN) involves a simulation day-to-day ground-level ozone forecasting over a range of the human brain to model and predict the dynamics of of cities in USA using weather parameters as predictor an unknown system from sample sets of input-output data and, his ANN model gave somewhat better prediction without explicitly determining the underlying than multiple regressions. Several others have reported relationships. Mathematical neurons are constructed at similar efforts in various settings [8 -11]. Their works various layers as outlined in Figure 1.0; to process data show that artificial intelligence techniques such as ANN hence a neural network system has a good pattern play a great role in handling socio-economic and recognition capability. This is achieved via the use of environment issues quite well. appropriate activation functions f, in the basic input (x) – output(y) relationship given by 3. METHODOLOGY

Flaring and emissions data were collected from major Nigerian Oil & Gas firms and the Federal Environmental Protection Agency. Firstly, the mean and standard deviation of the training set are normalized. The normalization is such that the dataset would have zero mean and unity standard deviation. Due to the redundancy in the data set, we eliminated those principal components that contributed less than 0.001% to the total variation in the data set. The parameters were given with 264 sample cases which were partitioned into training set and testing set. For input, we considered stack height and flame length, both given in meters, the temperature ratio, the gas flow rate in meter cube per second, the H2S in solution gas in kilograms per mole and the combustion efficiency. The target output data were the percentage concentrations of Benzene, Toluene, Carbon, C3HC, C4HC, Styrene, Acetylene and Xylene.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

We recognized the fact that the gas flare problem is too Table 2: Regression analysis between the network complex to be solved with a single layer neural network. response and the corresponding target Multilayer networks have been proven to approximate almost any function as long as there are enough neurons Output Slope Intercept Variation in the hidden layer. Since our model involves 6 inputs Benzene 0.6 24.7 0.5292 and 8 outputs, we utilized the multilayer perception Toulene 0.9 13.7 0.6571 implemented in Matlab as a feed-forward neural network Carbon 0.6 19.7 0.4720 architecture with back propagation. It has 2 hidden layers C3HC 0.3 59.9 0.2522 containing 200 neurons each. An MLP network is C4HC 0.3 2.9 0.2393 adapted because this model was developed for multi- Styrene 0.4 23.7 0.3050 input and multi-output relationships, in which Acetylene 0.3 5.9 0.2798 optimization is sought for error correction due to the non- Xylene 0.4 2.3 0.3073 linearity involved. For data pre-processing, the mean and standard deviation of the training set are normalized. The normalization is such that the dataset would have zero The results as provided in Table 2 show that our ANN mean and unity standard deviation. Due to the performed best for Toulene, Benzene and Carbon with redundancy in the data set, we eliminated those principal degraded accuracy for Acetylene, C HC and C HC. The components that contributed less than 0.001% to the total 3 4 plots of the ANN simulation are illustrated in Figures 2 to variation in the data set. This technique has 3 effects; 9. (a) It orthogonalizes components of input vectors

so that they uncorrolate into each other. (b) It orders the resulting orthogonal components so that those with the largest variation come first. (c) It eliminates those components that contribute the least to the variation in data set.

4. DISCUSSION OF SIMULATION RESULTS

The results were obtained after a series of trial and error on the number of neurons per layer and the number of hidden layers. The described configuration was chosen after several performance trials. The appropriate connection weights and neuron thresholds were found so that the network produced appropriate outputs for each input in its training data. The network was tested on its training data and on new data.

From the trials and analysis;

M is the slope of the best linear regression relating targets to network outputs.

B is the intercept of the best linear regression relating targets to network outputs. If we had a perfect fit (outputs exactly equal to targets), the slope would be 1, and the y- Figure 2: Regression analysis between the network intercept would be 0. response 1 and the corresponding target 1

R is a measure of how well the variation in the output is explained by the targets. If this number is equal to 1, then there is perfect correlation between targets and outputs.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 3: Regression analysis between the network Figure 5: Regression analysis between the network response 2 and the corresponding target 2 response 4 and the corresponding target 4

Figure 4: Regression analysis between the network response 3 and the corresponding target 3 Figure 6: Regression analysis between the network response 5 and the corresponding target 5

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 7: Regression analysis between the network response 6 and the corresponding target 6 Figure 9: Regression analysis between the network

response 8 and the corresponding target 8

5. CONCLUSIONS The relative concentrations of flare pollutants are predicted by the ANN model. The simulation exercise shows some agreement with test data. This could be used for data quality control or reconstruction of missing data by relevant agencies. The problems caused by gas flaring can be further reduced if proper planning is done with the aid of accurate forecasting techniques like artificial neural networks. Such prediction could aid the formulation of polices to check gas flaring such as permissible levels of pollutant constituents over time so as to check hazardous exposures and environmental degradation. The research study has shown that more the system parameters combinations such as number of hidden layers, nodes, trials, etc. need to be investigated for optimal performance. Future work could consider comparison of our results/run-time with those obtained by other techniques.

Figure 8: Regression analysis between the network ACKNOWLEDMENTS The author acknowledges the field work done by his student; Mary-Yakndara Effiom response 7 and the corresponding target 7

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFERENCES [1] The Guardian Newspapers, 2008. “Nigeria Loses [9] C. Marzaban, and G. J. Stumptf, 1996. “A N6b Daily To Gas Flaring,” Aug 5, 2008 pp 1-4, Neural Network for Tornado Prediction Guardian Newspapers Limited, Lagos, Nigeria. Based on Doppler Radar- Derived Attributes,” Journal of Applied Meteorology, 35:615-626. [2] ThisDay Oil Report, 2010. “Nigeria and other oil- producing countries: a comparative study,” Leaders [10] A. M. Russel, M. S. Bergin, S. M. McBride, L. & Company Ltd, Apapa, Lagos. McNair, Y. Yang, W. R. Stockwell and B. Croes, 1995. “Urban ozone Control [3] World Bank Report, 2007. “Reported Flaring Data and Atmospheric Reactivity of Organic for 2004-2005.” Assessed online at Gases,” Science, 269:491-495. http://web.worldbank.org/WBSITE/EXTERNAL/T OPICS/EXTOGMC/EXTGGFR/0,,contentMDK:2 [11] G. Sola, and A. M. Sola, 1999. “Ozone Indices 1348978~isCURL:Y~menuPK:2912270~pagePK: Based On Simple Meterological Parameters: 64168445~piPK:64168309~theSitePK:578069,00. Potentials and Limitation Of Regression html on July 14, 2011 and Neural Network Models,” Atmospheric Environment, 33:4299-4307. [4] E. O. Nwaichi and M. A. Uzazobona, 2011. “Estimation of CO2 level due to gas flaring in the Niger Delta,” Research Journal of Environmental Sciences, 5(6):565-572, Academic Journal Inc Author’s Brief

[5] F. Benvenuto and A. Marani, 2000. “Nowcasting

of Urban Air Pollutants by Neural Networks,” Olumuyiwa S. Asaolu is Nuovo Cimento, 23C, 567-586. currently a senior lecturer in Systems Engineering at the [6] G. Nunnair, A. F. M. Nucifora and C. Randieri, University of Lagos, Nigeria. 1998. “The Application of Neural Techniques to He has a PhD in Engineering the Modeling of Time-Series of Atmospheric Analysis and specializes in Pollution Data,” Ecological Modeling, 111:187- Artificial Intelligence. He is a recipient of several scholarly 205. awards.

[7] A. C. Comrie, 1997. “Comparing Neural Networks and Regression Models for Ozone Forecasting.” Journal of Air and Waste Management Association, 47:653-662.

[8] P.A. Blamire, 1996. “The Influence of Relative

Sample Size in Training Artificial Neural Networks,” International Journal of Remote Sensing, 17:223-230.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Empirical Evaluation of Customers’ Use of Electronic Banking Systems in Nigeria

Onyedimekwu Okechi AfriHUB Nig. Ltd ICT Solutions for Africa Federal college of Education (Technical) Omoku, Rivers State, Nigeria. [email protected]

Oruan Memoye Kepeghom Department of Computer Science Federal College of Education Omoku, River State, Nigeria [email protected]

ABSTRACT Electronic banking systems enable customers to access banking services through intelligent electronic devices such as Computers (Internet banking), Personal Data Assistants (PDAs), Mobile Phones (Mobile banking & Mobile Money), Point of sales Terminals (PoS), Automated Teller Machines (ATMs), and Debit Cards etc. This research focuses on empirically evaluating customers’ use of electronic banking systems. DeLone and McLean Information System Success model (2003) was employed as a conceptual framework. The survey instrument employed involved design and administration of 240 questionnaires within Omoku town in Rivers state. 14 returned questionnaires were rejected due to wrong filling. A total of 220 questionnaires were analyzed which represents 91.7%. The result of this research shows that among all e-Banking systems, ATM has the highest level of usage. The percentage of respondents who claimed to always use the various forms of e-banking systems is as follows: ATMs (22.7), PoS (6.4%), Internet Banking (7.3%), Mobile Banking (10.5%), Mobile Money (8.7%), MasterCard (11.0%), and Web Merchants (5.5%). Correlation analysis of the hypothesis variables indicates the following: System Quality and Continuance Intention( ᵝ = .421,ρ = .000 ), Information Quality and Continuance Intention( ᵝ = .437,ρ = .000), Service quality and Customers’ Satisfaction( ᵝ = -0.097,ρ =. 150). Most bank customers were not satisfied with ATM service Quality in terms of how banks handle their customer complaints, functionality of the ATM and long queues in using the ATM. Bank customers should be well informed on how to use all forms of e-banking systems for their financial transactions.

Keywords- Electronic Banking Systems, ATM debit Cards, Information systems, System quality and User Satisfaction.

African Journal of Computing & ICT Reference Format: Onyedimekwu Okechi & Oruan Memoye Kepeghom (2013). Empirical Evaluation of Customers’ Use of Electronic Banking Systems in Nigeria. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 7-20

1. BACKGROUND 1.1 Theoretical Framework Using D & M Information A recent research by Adesina and Ayo [1] found out that System Success Model “all members of the Nigeria banking industry have Evaluations of information systems can be performed engaged the use of Information and Communication through different approaches and methodologies and Technology (ICT) as a platform for effective and consequently evaluations aim to fulfill different kinds of efficient means of conducting financial purposes and produce different kinds of results [2]. transactions”(p.2). Electronic banking systems enable Information Systems (IS) like Electronic Banking customers to access banking services through intelligent Systems are developed using information technology (IT) electronic devices such as Computers (Internet banking), to facilitate banking services. Researchers have derived a Personal Data Assistants (PDAs), Mobile Phones (Mobile number of models to explain what makes some IS banking & Mobile Money), Point of Sales Terminals ‘successful’. Davis’s Technology Acceptance Model (PoS), and Automated Teller Machines (ATMs), Debit (TAM) [3] used the Theory of Reasoned Action (TRA) Cards etc. and Theory of Planned Behavior (TPB) to explain why some IS is more readily accepted by users than others.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Acceptance, however, is not equivalent to success, To address this problem DeLone and McLean [6] although acceptance of an information system is a performed a review of the research published during the necessary precondition to success [4]. The pre-adoption period 1981–1987, and created taxonomy of IS success intention to use measure from the TAM model is not as based upon this review. In their 1992 paper, they suitable to use for an evaluative purpose [5]. Early identified six variables or components of IS success: attempts to define information system success were ill- system quality, information quality, use, user defined due to the complex, interdependent, and multi- satisfaction, individual impact, and organizational dimensional nature of IS success. impact. However, these six variables are not independent success measures, but are interdependent variables. Figure 1 shows this original IS success model [6].

System Use Quality

Individual Organizational Impact Impact

Information User quality Satisfaction

Figure 1: DeLone and McLean IS success model (1992).

Shortly after the publication of the D&M success model, The updated model is shown in Figure 2. The dimensions IS researchers began proposing modifications to this of success based on [4] include: model. Seddon and Kiew (1996) studied a portion of the IS success model (i.e. system quality, information System quality – the desirable characteristics of an quality, use, and user satisfaction). In their evaluation, information system. For example: ease of use, Intention they modified the construct, use, because they to Use ,flexibility, system reliability, and ease of “conjectured that the underlying success construct that learning, as well as system features of intuitiveness, researchers have been trying to tap is Usefulness, not sophistication, and response times. Use” (p. 93). Seddon and Kiew’s concept of usefulness is equivalent to the idea of perceived usefulness in TAM by Information quality – the desirable characteristics of the [3]. They argued that, for voluntary systems, use is an system outputs; that is, management reports and Web appropriate measure; however, if system use is pages. For example: relevance, understandability, mandatory, usefulness is a better measure of IS success accuracy, conciseness, completeness, understandability, than use. DeLone and McLean [4] responded that, even currency, timeliness, and usability. in mandatory systems, there can still be considerable variability of use and therefore the variable use deserves Service quality – the quality of the support that system to be retained. Researchers have also suggested that users receive from the IS department and IT support service quality be added to the DeLone and McLean personnel. For example: responsiveness, accuracy, (D&M) model. reliability, technical competence, and empathy of the Personnel staff. SERVQUAL, adapted from the field of Recognizing these proposed modifications to their model, Marketing, is a popular instrument for measuring IS D&M, in a follow-up work, reviewed empirical studies service quality Pitt et al. [7]. that had been performed during the years since 1992 and revised the original model accordingly [4]. In their Intention to use – the degree and manner in which staff updated model, they included the core dimensions of and customers utilize the capabilities of an information information quality, system quality and user system. For example: amount of use, frequency of use, satisfaction, as well as use/intentions to use, net nature of use, appropriateness of use, extent of use, and benefits and service quality. purpose of use. Bhattacherjee [8] makes the case for

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

assessing user intentions to continue using a system economy, since according to Lagsten and Goldkuhl [2] (continuance intention). Such a measure is more suited to “One major reason for doing evaluations of information post-adoption studies such as evaluating the success of an systems is to take actions based on the results of the information system. evaluation”(p.97).

User satisfaction – users’ level of satisfaction with any This work also seeks to validate DeLone and McLean information system. For example, the most widely used (2003) IS Success Model using ATM debit card as an multi-attribute instrument for measuring user information Electronic Banking System in Nigeria. ATM debit card satisfaction can be found [9]. User Satisfaction is the usage ranked far higher than other Electronic Banking most general perceptual measure of information systems Systems according to recent findings by Central Bank success [10]. Nigeria [13] and Intermarc [14].

Net benefits – the extent to which IS are contributing to 1.4 Research Questions and Hypotheses the success of individuals, groups, organizations, Many users of Electronic Banking Systems (EBS) in industries, and nations. For example: improved decision- Nigeria are very Skeptical of the system and the making, improved productivity, increased sales, cost following questions are reductions, improved profits, market efficiency, consumer welfare, creation of jobs, and economic Q1. What is the system quality of an Electronic development. Brynjolfsson et al. [11] have used Banking System in terms of its security, production economics to measure the positive impact of reliability, Ease of Use and Availability? IT investments on firm-level productivity. The practical This research proposes the following application of the D&M model is naturally dependent on hypotheses based on EBS quality: the organizational context. DeLone and McLean [4] stated that the researcher wanting to apply the D&M H1: The System Quality of an Electronic Banking model must have an understanding of the information System has a positive effect on the system and organization under study. This will determine Customers’ continuance Intention to use the types of measures used for each success dimension. it. The selection of success dimensions and specific metrics depend on the nature and purpose of the system(s) being H2: The System Quality of an Electronic Banking evaluated [4]. System has a positive effect on the Customer’s Satisfaction. 1.2 Statement of Research Problem Nigerian banks have been investing huge sums of money Q.2 What is the Information Quality content of an in ICT and e-Banking systems most especially in the Electronic Banking system in terms of deployment of ATM and issuance of ATM debit cards account balance enquiring, printing towards measuring up with global standards. Customers transaction record etc? Based on this metric, of banks today are no longer only concerned about safety the following hypothesis is proposed: of their funds and security of ATM transactions but customers demand efficient, fast and convenient services. H3: The Information Quality content of an Customers want a Bank that will always have ATM Electronic Banking system has a positive services, reduced queue, reliable and secure internet and effect on the Customers’ continuance mobile services, support their business goals for instance; Intention to use it. businessmen want to travel without carryout cash for security reasons. They want to be able to check their H4: The Information Quality content of an balance online, find out if a cheque is cleared, transfer Electronic Banking system has a positive effect funds among accounts and even want to download on the Customers’ satisfaction. transaction records into their own computer at work or home. Evaluating the success level of these Electronic Q3. What is the Service quality of an Electronic Banking systems especially ATM cards usage from the Banking system in terms of ATM customers’ perspective will boost their confidence and functionality, technical support during usage, willingness to continue using these systems. Secure Service, average time spent on usage?

1.3 Purpose of Study H5: The Service Quality of an Electronic Banking Electronic Banking Services like ATM and debit cards System has a positive effect on the witness a tremendous revolution with the introduction of Customers’ continuance Intention to use Guideline on Electronic Banking by the Central Bank of it. Nigeria [12].This research work seeks to empirically evaluate the success of Electronic Banking systems in Nigeria, and to access customers readiness for cashless

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

H6: The Service Quality of an Electronic Banking System has a positive effect on the Customers’ Satisfaction Intention to use (modified to Continuance Intention according to [8]) Q4. How does a customers’ continuance intention This is the degree and manner in which staff and to use EBS affect customers’ satisfaction customers utilize the capabilities of an information and Net Benefit of using the system? system. For example: amount of use, frequency of use, This research proposes the following nature of use, appropriateness of use, extent of use, and hypotheses based on continuance intention to purpose of use. Bhattacherjee [8] makes the case for use an electronic banking system. assessing user intentions to continue using a system (continuance intention). Such a measure is more suited to post-adoption studies such as evaluating the success of an information system like ATM debit cards.

System Continuance Intention Quality

Net Information Benefit Quality

Service User Quality Satisfaction

Fig. 2 Electronic Banking Systems’ Research Hypotheses Model

H7: Customers’ continuance Intention to use H10: Customers’ Satisfaction of an Electronic Electronic Banking System has a positive effect Banking System has a positive effect on the on the Customers’ Satisfaction. Customers’Net Benefit.

H8: Customers’ continuance Intention to use 1.5 Limitations of the Study Electronic Banking System has a positive effect Evaluation of Electronic Banking Systems is a complex on the Customers’ Net Benefit. work since there are many systems to be considered. This research work limited the evaluation to customers’ Use of Q5. How does customers’ satisfaction of an ATM debit card using DeLone & McLean IS Success electronic banking system affect Net benefit Model. The survey questionnaires were administered in and continuance intention of using it? Omoku Town only due to limited financial resources. Also questions were only administered to customers of H9: Customers’ Satisfaction of an Electronic banks who use ATM debit cards or are intending to use Banking System has a positive effect on the it. Commercial banks’ staffs’ perspectives were not Customers’ continuance Intention to use it. considered.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2. RELATED LITERATURE Banking system use it because it is convenient, easy to 2.1. Introduction use, time saving and appropriate for their transaction Electronic Banking System is an innovative service needs. Also the network security and the security of the delivery mode that offers diversified financial services system in terms of privacy are the major concerns of like cash withdrawal, funds transfer, cash deposits, the users and constitute hindrance to intending users” payment of utility and credit card bills, cheque book (p.2). Thomas et al. [16] carried out a research on the requests, and other financial enquiries. In Nigeria, Importance and Performance of Various Factors ATM was conventionally introduced as an electronic Considered In the Electronic Banking Services and delivery channel in 1989, and was first installed by concludes that “secure service is the most important National Cash Registers (NCR) for the defunct Societe dimension, followed by convenient location of ATM, Generale Bank of Nigeria (SGBN) in the same year. efficiency (not need to wait), ability to set up accounts Since its introduction, many Nigerian banks have so that the customer can perform transactions installed ATM in response to the changing nature of immediately, accuracy of records, user friendly, ease of modern banking operations. Until 2003, a small use, complaint satisfaction, accurate transactions and number of banks operated their own propriety ATM operation in 24 hours” (p.151). Available data on fleets .The main shared ATM network in Nigeria, various e-payment channels from the Central Bank of InterSwitch, began operations in 2003 with 5 ATMs Nigeria Economic Report for the first half of 2011 from United Bank for Africa (UBA) and First Bank of revealed that “ATM remained the most patronized, Nigeria (FBN).[15]. accounting for 98.09 percent with a transaction worth of N764.14 billion. The number of ATM deployed in 2.2. Views on Electronic Banking the system stood at 9,443 with over 30 million ATM Recent research work by Adesina and Ayo [1] found card holders at the end of June 2011(p.6) [13] out that “Banks’ customers who are active users of e-

Fig. 3 Percentage of Volume and Value of E-payment channels, Half year 2011(CBN, 2011)

2.3. Electronic Banking Systems 2.3.3 Mobile Money allows users to create an e-wallet 2.3.1 Telephone Banking is a service provided by a for storing funds on their phone. Once value is stored on financial institution, which allows its customers to your mobile phone, you can use it to pay for goods and perform some banking transactions over the telephone. services at merchant locations that support mobile Most telephone banking services use an automated phone money. It is at the core of the CBN’s cashless policy. The answering system with phone keypad response or voice philosophy behind mobile money is that most Nigerians recognition capability. now have mobile phones (not as many have bank accounts). So, if we can have an easy to use electronic 2.3.2 Mobile Banking (also known as M-Banking, payment solution that enables people pay for goods and mbanking) is a term used for performing balance checks, services with their mobile phone we will achieve a account transactions, payments, credit applications and cashless or cash lite economy faster.Your e-wallet can be other banking transactions through a device such as a funded via authorized agents of your Mobile money mobile phone or Personal Digital Assistant (PDA). The service, partner banks and networks of your mobile earliest mobile banking services were offered over Small money service, transfers from your ATM/Debit cards, or Message Service (SMS). any other funding method offered by your service provider.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Once funded you can securely and conveniently use your Different ATM service providers are also interconnected mobile money to send money to family and friends, buy so you do not need to worry which company services a airtime of any network, Pay bills like DSTV, Hi TV, particular ATM machine. InterSwitch for example MyTV, PHCN bills, etc. The Central Bank of Nigeria supports Verve, Visa and MasterCard on her ATM (CBN) recently issued operating licenses to 11 mobile machines and vice versa. money firms, namely: Fortis Mobile Money, UBA/Afripay, GTBank Mobile Money, Pagatech, and 2.3.7 Smart Card (chip card or integrated circuit eTranzact, Monetise, Eartholeum, Paycom, FET, card) is any pocket-sized card with embedded integrated Ecobank and Kudi. The operating license allows the circuits which can process data. The Verve card is the companies to provide products such as electronic new Interswitch debit card now with the Chip and payments through mobile phones [17]. PIN. Verve is the new name of the more secure and convenient Interswitch card which with the introduction 2.3.4 Online Banking (or E-banking) of chip and PIN makes transactions safer and with Quick This allow customers of a financial institution to conduct teller services, adds convenience to everyday life. Verve financial transactions on a secure website operated by the is issued by your bank and powered by Interswitch. The institution, which can be a retail or virtual bank, credit Verve card is the first and only chip card accepted on all union or building society. To access a financial available payment channels in Nigeria. The Verve card is institution's online banking facility, a customer having the only card that allows you to conveniently pay for personal Internet access must register with the institution goods and services on all ATMs, POS, Web, Mobile, for the service, and set up some password (under various Kiosk, PC POS, Voice and Bank Branch connected to the names) for customer verification. Interswitch network. The Verve card is one of the most secure Chip and PIN card. The chip technology 2.3.5 Point-of-Sale (PoS) terminal guarantees that information stored is not accessible to This is an electronic device that is used for verifying and unauthorized persons [18]. processing credit card transactions. Typically connected via highly reliable telephone wired connections, they 2.3.8 Web Merchants. Web merchants are those require rapid dial up time, low power and reliable organizations that conduct transaction via their websites. performance. A Retail Point of Sales system typically They make it possible for people to buy goods or render includes a computer, monitor, cash drawer, receipt services to people via their websites. Examples of these printer, customer display and a barcode scanner, and the web merchants are www.quickteller.com, majority of retail POS systems also include a debit/credit Virtualkard.com and Naira.com. They offer prepaid card reader. It can also include a weight scale, integrated services and charges are based on prevailing currencies credit card processing system, a signature capture device exchange rate. Online buyer purchases the card based on and a customer pin pad device. More and more POS the amount of the product(s) in dollars that he wants to monitors use touch-screen technology for ease of use and buy and uses the card information to make his online a computer is built in to the monitor chassis for what is purchases. Another payment method is direct payment to referred to as an all-in-one unit. the seller’s bank. This is an offline method. Some of the companies using this method are syskay.com and 2.3.6 Automated Teller Machine (ATM). Signonafrica.com. Both are web hosting companies. It is An ATM device allows a bank customer to withdraw required that a customer pays into the bank account of the cash from his account via cash dispenser (Machine), and seller or service provider. A proof of payment is required, the account is debited immediately. A fundamental sometimes the teller is scanned and sent via email as advantage is that it needs not to be located within the proof of payment before the service is rendered, and a banking premises. It is usually in stores, shopping malls, process called Non-repudiation the customer makes a fuel stations etc. On most modern ATMs, the customer is phone call [19]. identified by inserting a plastic ATM card with a magnetic stripe or a plastic smart card with a chip, that 2.4. The Emerging Issues in E-Banking Systems contains a unique card number and some security 2.4.1 Threats to Electronic Banking Systems information such as an expiration date. A Miami businessman is suing his bank for the loss of $90,000. He claims that, in February 2005, this money Authentication is provided by the customer through was stolen from his online bank account via an entering a personal identification number (PIN). unauthorized transaction [20]. Investigations have InterSwitch, VPay, ETranzact, and QuickCash are some revealed that the businessman’s computer was infected of the leaders in ATM deployment in Nigeria. with a Trojan capable of logging keystrokes [21], InterSwitch today has all banks in the country connected including his full account details. It is likely that the theft to her network. This actually makes it possible to use of this information was the trigger that led to the their cards in all bank branches nationwide and in almost unapproved transaction to a foreign bank account. all machines.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

So far, the businessman’s bank has refused to compensate Why the Cash Policy? for his loss. The number of malicious applications The new cash policy was introduced for a number of key targeting online banking transactions has increased reasons, including: dramatically in recent years and the biggest threat to 1. To drive development and modernization of online banking is still malicious code executed carelessly our payment system in line with Nigeria’s on the end-user’s computer [21]. The attackers tend to vision 2020 goal of being amongst the top 20 target the weakest link. Once the attacker has control economies by the year 2020. An efficient and over a user’s computer, he or she can modify the modern payment system is positively correlated information flow to his or her advantage. ATMs have with economic development, and is a key contributed to the alarming rate of fraud in the Nigerian enabler for economic growth. banking industry [22]. 2. To reduce the cost of banking services (including cost of credit) and drive financial 2.4.2. The Regulatory Challenges inclusion by providing more efficient Recognizing that electronic banking and payments transaction options and greater reach. services are still at the early stages of development in 3. To improve the effectiveness of monetary Nigeria, Central Bank of Nigeria released it’s guidelines policy in managing inflation and driving on electronic banking in Nigeria [12]. According to the economic growth. guideline, Banks will be considered liable for fraud arising from card skimming and counterfeiting except In addition, the cash policy aims to curb some of the where it is proven that the merchant is negligent. negative consequences associated with the high usage of However, the cardholder will be liable for frauds arising physical cash in the economy, including: from PIN misuse. External devices such as Automated Teller Machines (ATMs), Personal Computers, (PC’s) at High cost of cash: There is a high cost of cash along the remote branches, kiosks, etc. permanently connected to value chain - from the CBN & the banks, to corporations the bank’s network and passing through the firewall must and traders; everyone bears the high costs associated with at the minimum address issues relating to non- volume cash handling. repudiation, data integrity and confidentiality. High risk of using cash: Cash encourages robberies and The CBN guideline addresses the issues of security by other cash-related crimes. It also can lead to financial loss stating that “Banks may consider authentication via in the case of fire and flooding incidents. Media Access Control (MAC) address in addition to other methods. Adopt the chip (smart card) technology as High subsidy: CBN analysis showed that only 10 percent the standard. Banks must ensure that the Internet Service of daily banking transactions are above 150k, but the 10 Provider (ISP) has implemented a firewall to protect the percent account for majority of the high value bank’s Web site where outsourced” (p.3). Section 1.4.8 transactions. This suggests that the entire banking of the guideline states that “Internet Service Providers population subsidizes the costs that the tiny minority 10 (ISPs) should exercise due diligence to ensure that only percent incurs in terms of high cash usage. websites of financial institutions duly licensed by the CBN are hosted on their servers. ISPs that host Informal Economy: High cash usage results in a lot of unlicensed financial institutions would therefore be held money outside the formal economy, thus limiting the liable for all acts committed through the hosted effectiveness of monetary policy in managing inflation websites.” and encouraging economic growth.

2.4.3. Electronic Banking Systems and Cashless Inefficiency & Corruption: High cash usage enables Economy corruption, leakages and money laundering, amongst The Central Bank of Nigeria (CBN) has introduced a new other cash-related fraudulent activities [23] policy on cash-based transactions which stipulates a ‘cash handling charge’ on daily cash withdrawals or cash 3 RESEARCH METHODOLOGIES deposits that exceed N500, 000 for Individuals and N3, 3.1 Introduction 000,000 for corporate bodies. The new policy on cash- The methodology employed in this study was positivistic, based transactions (withdrawals & deposits) in banks, quantitative and hypothetic-deductive. Hypotheses were aims at reducing (NOT ELIMINATING) the amount of derived from the extant literature on Information physical cash (coins and notes) circulating in the Systems’ Evaluation using D & M IS Success Model. economy, and encouraging more electronic-based transactions (payments for goods, services, transfers, through ATMs, PoS, Website, Mobile Phones etc.)[23].

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

3.2 Research Design The questionnaire items were adopted from the following Questionnaire is the survey instrument used in this prior studies [1] [5]. Bank customers were asked to research. The research design was divided into two indicate their perception on the use of ATM cards as an sections. The first section consists of 5 questions on E-banking system using the Five-Point Likert’s scale demographic profile and 7 questions on the use of various having the ratings of “strongly disagree” (1) and e-banking systems. The second section consists of the “strongly agree” (5). evaluating factors (items) and 20 questions which test the customers’ usage of ATM debit card as shown below. 3.3 Research population and Sampling Procedure The research population of study includes bank Table1: Hypothesis Variables and numbers of customers who are using e-banking services, staffs and research questions students of federal college of education, Omoku, NYSC Variables Numbers of members, and anybody who has a bank account. A criteria standard to define the “valid questionnaire” is set Items(questions) as follows: a questionnaire having more than 10 items Systems quality (SQ) of ATM 7 clicked continuously in the same score was considered to be invalid; if up to 5 questions were not answered, it is debit card(s) considered invalid. Random sampling of questionnaires Information quality (IQ) 3 to bank customers in Omoku was used.

Service Quality (SQ) 4 3.4 Data Collection (procedure) Continuance Intention to use 3 Primary and secondary data sources were used in this work. The primary data source is the questionnaire while ATM Cards (CI) secondary source is based on Central Bank of Nigeria’s Customers’ Satisfaction (CS) 2 half year 2011 report [13], Intermarc consulting report [14] and others. 240 questionnaires were distributed and Net Benefit of using ATM cards 1 234 were collected from the respondents from July 02, (NB) 2012 to August 10, 2012. 14 returned questionnaires were rejected due to incomplete or wrong filling. A total of 220 questionnaires were therefore analyzed which represents 91.7%.

Table 2: Demographic profile and e-Banking usage of respondents

Gender Frequency Valid percentage Male 129 58.9 Female 90 41.1 Age Range Frequency Valid percentage Below 20 Yrs 32 14.5 20-30 104 47.3 31-40 67 30.5 41-50 16 7.3 51-60 0 0.0 Above 60 Yrs 1 .5 Occupation Frequency Valid percentage Student 65 29.5 Civil Service 59 26.8 Private Job 34 15.5 Trading 10 4.5 Self-Employed 24 10.9 Others 28 12.8 Educational Qualification Frequency Valid percentage SSCE/WAEC 40 18.2 NCE 54 24.5 OND 11 5.0 HND 30 13.6 First Degree 75 34.1 Masters 7 3.2 PhD 2 .9 Monthly Income Frequency Valid percentage Below #20,000 84 38.2 20,000-35,000 45 20.5 40,000-65,000 27 12.3 70,000-100,000 16 7.3 Above 100,000 18 8.2 No Monthly Pay 30 13.6

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4. RESULTS, ANALYSIS AND DISCUSSION

4.1. Introduction The collected data were analyzed based on descriptive statistics (frequency and percentage) and correlation analyses using the statistical package for social sciences (SPSS) version 18.

4.2. Research Questions/Hypotheses Analysis and Results In this section the demographic profile and e-banking usage of the respondents will be stated. Also reliability analysis of the variables as well as the descriptive statistics is given.

4.2.1 Demographic profile and e-Banking usage Two hundred and twenty (220) questionnaires were analyzed. The demographic profile of the respondents is presented in the table below.

Table 3: Electronic Banking Systems Usage of respondents ATM PoS Internet Mobile Mobile Master Web Banking Banking Money Card Merchants

Freq % Freq % Freq % Freq % Freq % Freq % Freq %

I don’t use 65 29.5 165 75.0 168 77.1 153 69.5 162 74.3 147 67.1 175 80.3 it

Once per 62 28.2 18 8.2 16 7.3 23 10.5 24 11.0 23 10.5 17 7.8 month

Once per 30 13.6 16 7.3 7 3.2 13 5.9 7 3.2 16 7.3 4 1.8 week

Twice per 11 5.0 4 1.8 11 5.0 8 3.6 6 2.8 9 4.1 9 4.1 week

Always 50 22.7 14 6.4 16 7.3 23 10.5 19 8.7 24 11.0 12 5.5

Figure 4: ATM debit Cards Usage of respondents

4.2.2 Reliability Analysis of Variables In order to assess reliability of the variables, the Cronbach Alpha was calculated for each variable. The lowest Cronbach Alpha was 0.405 for Customer Satisfaction and the highest was 0.534 for Information Quality, thus demonstrating that all measures exhibited reliability.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Table 4: Reliability Analysis of Variables

Variable Number of Items Cronbach alpha Systems quality (SQ) of ATM debit card(s) 7 0.486 Information quality (IQ) 3 0.534 Service Quality (SQ) 4 0.493 Continuance Intention to use ATM Cards (CI) 3 0.468 Customers’ Satisfaction (CS) 2 0.405 Net Benefit of using ATM cards (NB) 1 N/A

4.2.3 Correlation Analysis of Variables The table below shows the correlation analysis of the 6 variables and their mean value from SPSS output.

Table 5: Correlation Analysis of Variables

SYSTEM INFORMATION SERVICE CONTINUANCE CUSTOMER NET QUALITY QUALITY QUALITY INTENTION SATISFACTION BENEFIT MEAN SYSTEM Pearson 1 .445** -.049 .421** .376** .224** 3.342 QUALITY Correlation Sig. (2- .000 .467 .000 .000 .001

tailed) N 220 220 220 220 220 220 INFORMATION Pearson .445** 1 .056 .437** .317** .146* 3.395 QUALITY Correlation Sig. (2- .000 .407 .000 .000 .031

tailed) N 220 220 220 220 220 220 SERVICE Pearson -.049 .056 1 .138* -.097 -.069 3.364 QUALITY Correlation Sig. (2- .467 .407 .041 .150 .311

tailed) N 220 220 220 220 220 220 CONTINUANCE Pearson .421** .437** .138* 1 .424** .275** 3.449 INTENTION Correlation Sig. (2- .000 .000 .041 .000 .000

tailed) N 220 220 220 220 220 220 CUSTOMER Pearson .376** .317** -.097 .424** 1 .433** 3.318 SATISFACTION Correlation Sig. (2- .000 .000 .150 .000 .000

tailed) N 220 220 220 220 220 220 NET BENEFIT Pearson .224** .146* -.069 .275** .433** 1 3.636 Correlation Sig. (2- .001 .031 .311 .000 .000

tailed) N 220 220 220 220 220 220

**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4.2.4 Hypothesis testing Hypotheses were supported if the p value is less than 0.05.

Table 6: Hypothesis Testing Hypothesis Independent Variable Dependent Variable Beta Value P level (p<0.05) Hypothesis Supported

H1 System Quality Continuance Intention .421 .000 YES

H2 System Quality Customer’s Satisfaction .376 .000 YES

H3 Information Quality Continuance Intention .437 .000 YES

H4 Information Quality Customer’s Satisfaction .317 .000 YES

H5 Service quality Continuance Intention .138 .041 YES

H6 Service quality Customer’s Satisfaction -0.097 .150 NO

H7 Continuance Intention Customer’s Satisfaction .424 .000 YES

H8 Continuance Intention Net Benefit .275 .000 YES

H9 Customer’s Satisfaction Continuance Intention .424 .000 YES

H10 Customer’s Satisfaction Net Benefit .433 .000 YES

4.3. Discussion of Results The mean of Service Quality shows that most Table 2 shows that 58.9 percent of the respondents respondents were not satisfied by ATM service Quality were males while 41.1 percent were females. 47.3 in terms of how banks handle their customer percent of them were aged between 20-30 years. 29.5 complaints, functionality of the ATM, success of ATM percent are students while 26.8 percent are Civil cards transactions, and long queues in using the ATM Servants. Most of the respondents had bachelor degree which was tested via the questionnaire. This means that (34.1%). Income earners of below #20,000 monthly if banks can increase their Service quality based on the had the highest percentage of 38.2. E-Banking variables tested in this research work, then customers’ Systems’ usage of the respondents was also satisfaction will increase leading to more people determined. Respondents were asked to indicate the willing to use the system. rate at which they use any of the various forms of e- Banking systems. The frequency distribution of their e- Banking system usage is illustrated in Table 3. Among The average value of Net Benefit which is 3.64 all e-Banking system, ATM has the highest level of indicates that ATM debits cards as an e-banking system usage in accordance with previous research by Adesina is beneficiary to most people. From the correlation and Ayo [1] and Central Bank of Nigeria Half year analysis of the variables in table 5, the strongest direct report [13]. 22.7% of the respondent claimed to always relationship existed between Information Quality (IQ) use ATM and 28.2% once per month use ATM. Web and Continuance Intention (CI) to use ATM services. Merchant Portal system was observed to be the least That is the vital, timely, secure and relevant used. 80.3% responded that they don’t use it for their e- information an ATM system has the higher the people’s banking transaction. Reliability Analysis of the intention and wiliness to continue it usage. Variables shows that the Cronbach Alpha value for System quality (SQ) is 0.486, Information quality is 0.534, Service Quality is 0.493, Continuance Intention is 0.468, and Customers’ Satisfaction (CS) is 0.405.

Table 5 shows the hypothesis testing, from the SPSS correlation analysis of the variable indicates that the entire hypothesis was supported except hypothesis 6.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

5. RESEARCH IMPLICATIONS AND REFERENCES CONCLUSION [1] Adesina, A.A. and Ayo, C.K (2010). An 5.1. Implication of the Study empirical investigation of the level of users’ The implication of the result of this study is that e- acceptance of E-Banking in Nigeria. Journal banking systems in Nigeria is still developing and of Internet Banking and Commerce, Vol. needs a lot of improvement from the commercial banks 15(1). Retrieved July 03, 2012, from deploying this services in terms of System quality, http://www.arraydev.com/commerce/jibc/ information quality and Service quality. The Central Bank of Nigeria, who is directly responsible for [2] Lagsten, J. and Goldkuhl, G. (2008). regulating e-banking services and its deployment and Interpretative IS evaluation: Results and other financial institutions must aggressively inform Uses. The Electronic Journal Information and encourage the public on how to use all the various Systems Evaluation Volume 11(2), pp. 97 – forms of e-banking systems.. 108. Retrieved July 09, 2012 from www.ejise.com 5.2. Conclusion This research work has used DeLone and McLean [3] Davis, F.D. (1989). Perceived usefulness, (2003) IS Success Model to show that most bank perceived ease of use, and user acceptance of customers will use e-banking systems more often if the information technology. MIS Quarterly system quality, information quality and service quality 13(3), 318–346. is improved. ATM debit cards usage ranked the highest while most bank customers do not know how to use [4] Delone, W.H. and Mclean E.R. (2003). The Web Merchant service and Mobile Money. The banks DeLone and McLean model of information should enlighten their customers about the gains of systems success: a ten-year update. Journal using e-banking systems and how to use them. of Management Information Systems 19(4), 9–30.

5.3. Recommendations [5] Brown, I. & Jayakody, R. (2008). B2C e- The researcher having critically examined the Commerce Success: a Test and Validation of responses of bank customers who use various forms of a Revised Conceptual Model. The Electronic e-banking systems as well as having read previous Journal Information Systems Evaluation works on e-banking systems and services here by Volume 11 (3), pp. 167 – 184. Retrieved from makes the following recommendations: http:// www.ejise.com 1) All money deposit banks in Nigeria should as a matter of urgency improve the service [6] Delone, W.H. and Mclean E.R. (1992). quality of their e-Banking Systems. Information systems success: the quest for 2) Bank customers should be encouraged and the dependent variable. Information Systems informed on how to use all forms of e- Research 3(1), 60–95. banking systems for their financial or business transactions, including using ATM [7] Pitt, L.F.; Watson, R.T.; and Kavan, C.B. debit card for payment purposes not just for (1995). Service quality: A measure of withdrawing money. information systems effectiveness. MIS 3) Further research work on customers’ Quarterly, 19(2), 173–188. perspective of the usage of e-banking systems by the CBN on a national base in other to [8] Bhattacherjee, A. (2001). An empirical cover the whole Nigerian state. This will analysis of the antecedents of electronic guide them properly when formulating e- commerce service continuance. Decisions banking policies and guidelines. Support Systems, Vol. 32(2), pp 201–214 4) Further work should be done on Biometric Authentication of ATM, for improving it [9] Ives, B.; Olsen M.; and Baroudi, J.J. The security. measurement of user information satisfaction. 5) Research work on the Design and Communications of the ACM, 26, 10 (1983), Construction of a unified (single) smart card- 785–793. based ATM debit card for all financial transactions.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

[10] Seddon, P.B. (1997). A Re-specification and [19] Adeyeye, M. (n.d). E-Commerce, business extension of the DeLone and McLean model methods and Evaluation of payment of IS success. Information Systems Research methods in Nigeria. The Electronic 8(3), 240–253 Journal Information Systems Evaluation Volume 11(1), pp. 1 – 6. Retrieved June 28, [11] Brynjolfsson, E., Hitt L.M. & Yang S. 2012, from http://www.ejise.com (2002). Intangible assets: how computers and organizational structure affect stock market [20] John Leyden (2005). Florida man sues bank valuations. Brookings Papers on Economic over $90k wire fraud. Retrieved Activity 1, p.137. June 26, 2012, from http://www.theregister.co.uk/2005/02/08/e [12] Central Bank of Nigeria (2003). Guideline on - banking_trojan_lawsuit/. Electronic Banking in Nigeria. www.cenbank.org/OUT/PUBLICATIONS/B [21] Symantec Security Response (July SD/2003/E-BANKING.PDF 2005).Threats to online banking: Virus Bulletin. [13] Central Bank of Nigeria (2011). Half year Retrieved June 27, 2012, from report (2011). Retrieved July 03 , 2012, from http://www.symantec.com/avcenter/refere http://www.cenbank.org/OUT/2011/PUBLIC nce /threats.t o.online.banking.pdf ATIONS/REPORTS/RSD/2011%20CBN%20HAL F%20YEAR%20REPORT.PDF [22] Chinedu, N.O., Chima B.O., and Emeka E.I. [14] Intermarc Consulting (2011): Status of (2012). Analysis of the negative effects of electronic payment industry in Nigeria: Card the Automated Teller Machine (ATM) as market survey application and production a channel for delivering banking machine. Retrieved July 03, 2012, from services in Nigeria. International Journal of http://www.intermarc- Business and Management, Vol. 7(7). ng.com/eventsdoc/coalition%20for%20epay Retrieved from ment%20presentation.pps http://www.ccsenet.org/ijbm

[15] Tope D. (2010). E-Banking Operations: The [23] Central Bank of Nigeria: Further Nigerian experience. Sales & Marketing clarifications on cash-less Lagos project. Manager NCR (Nigeria). Retrieved June 23, 2012, from http://www.cenbank.org/cashless/ [16] Thomas, O .O et al. Technology and Service Quality in the Banking Industry. African Journal of Business & Management Author’s Brief (AJBUMA), Vol. 1 (2010). Retrieved from http://www.aibuma.org/journal/index.htm Mr. Onyedimekwu Okechi [17] Microcapital Brief (September 7, 2011): obtained his B.Eng. (Elect/Elect) Central Bank of Nigeria Issues Mobile from Ahmadu Bello University, Money operating licenses. Retrieved July Zaria and his M.Sc. Information 4, 2012, from Technology from NOUN, Lagos. http://www.microcapital.org/microcapital- He is a Certified Internet Web brief- central-bank-of-nigeria-issues- Professional and a member of mobile-money- operating-licenses-to- IEEE Nigeria Section. He has affiliate-of-fortis-microfinance-bank-10- undergone professional practical computer training like other-companies/ CCNA, Web Development, Database Management, Computer Repairs and Maintenance. He is currently the [18] InterSwitch Limited (2011): Understanding Training Manager of AfriHUB Nig. Ltd (ICT Solutions the cashless economy. Retrieved June for Africa), FCE(T), Omoku, Rivers State. His research 25, 2012, from interest include: Information Systems analysis and http://www.interswitchng.com/#VERVE%20 design, E-Payment Systems, Web application DEBIT development with PHP/Mysql, Computer Networks and Use of ICT for Human Development. He can be contacted via [email protected], 08032667945..

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Mr. Oruan Memoye K. Obtained his B.Sc(Ed) Maths University of Nigeria Nsukka, PGD Computer Science and M.Sc. (comp. Sci) both from Uniport. He is an associate member of computer professionals registration council of Nigeria. He is currently a lecturer in computer science department,Federal College of education(T), Omoku. His research interest include Information Systems analysis and design, E-Payment Systems and Use of ICT for Human Development. He can be contacted via [email protected] 08037911980.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Design and Implementation of a Microcontroller Based Automatic Gate

O. Shoewu Department of Electronic and Computer Engineering Lagos State University Epe Campus, Lagos State, Nigeria. [email protected]

Segun O. Olatinwo Department of Computer Engineering Moshood Abiola Polytechnic, Abeokuta, Nigeria. Department of Computer Science and Engineering Ladoke Akintola University of Technology Ogbomoso, Nigeria. and [email protected]

ABSTRACT The common gate found almost everywhere has a lot of problem in term of operation, it is energy consuming, stressful and above all costly in term of paying for the man responsible for opening and closing of such gates, this then requires a fast solution. The microcontroller based automatic gate control is a better solution for the elimination of these problems caused by the manually controlled gates. The system monitors the gate as vehicles enter and exit the gate it is being mounted. The microcontroller based automatic gate senses any vehicle approaching as it cut across the path of the Infra red ray. After sensing this, the gate then automatically opens, wait for some time and closes after the time elapsed. The systems also work as an automatic lock, when the lock button is pressed that is when it is ON the gate does not open even if a vehicle cross the Infrared path. There is a special thing about this automatic gate I did in this project and that is it is totally controlled by both software and hardware, if there is any need for modification it can be easily modified by changing some part of the software since the microcontroller used can be reprogrammed, or if there is a hardware failure this can be changed and the system will be alright..

Keywords- Microcontroller, Gate Automation, PIC16f84, Infrared Technology.

African Journal of Computing & ICT Reference Format: O.Shoewu & O. Olatinwo (2013). Empirical Design and Implementation of a Microcontroller Based Automatic Gate. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 21-32

1. INTRODUCTION

Microcontroller based systems refine, extend or supplement human facilities and ability to observe, Microcontroller based automatic gate is an alternative to communicate, remember, calculate or reason and take a manually controlled gate which is laborious, frustrating, certain decision when necessary. In a search for making costly and energy consuming. Many are the devices Electronics Applications think, act and respond like which a microcontroller can be used in making some of Human, the proposed system was developed. The these are GSM phones, PDAs, Sound systems, Pumping proposed system attempts to make life more interesting machines, Robots e.t.c. The proposed system comprises by reducing unnecessary waste of man-power by of several component, the first is the sensors which detect employing microcontrollers. There are changes everyday, any vehicle near it and send a signal to another set of many things are being discovered due to technology component. It is interesting to note that this device can advancement, different devices are been discovered to perform some things like opening automatically, closing solve many of human’s problems. One of these problems automatically, lock up totally when the car park is filled can be solved using a microcontroller to control devices up and no vehicle can gain an entrance into the park. This thereby reducing the work of man. In addition to this work can be employed in public car parks, markets, development, human being is not resting in an effort to libraries, hotel, homes and anywhere that require the use find a solution to all of there problems and this project is of gates. It has many advantages over a manned gate in in no exception. the sense that it eliminate stress and salary of a gateman, also it can determine what to do next when a vehicle has come close to the gate.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2. SYSTEM OVERVIEW The proposed system on sensing an interruption from an infrared receiver, opens and closes the gates, also it can The proposed system is concerned with the design and deny an entrance or exit from the gate by just pressing of construction of a microcontroller based automatic gate. a button. These are the constituents of the automatic gate The proposed system monitors and controls a gate, we system; a sensor unit, a trigger circuitry, microcontroller employed a microcontroller that accepts data value from module, gate control unit and a power supply unit. The the interfacing circuit and take an instant decision. block diagram of the proposed system is below shown in figure 1.

SENSOR TRIGGER INTERFACING UNIT CICUITRY CIRCUITS. MICROCONTROLLER 16F84

GATE CONTROL UNIT DISPLAY UNIT

ELECTRIC GATE.

Fig 1: Block diagram of the proposed System.

The sensors in the system which are Infrared Input/Output controller (PIO). The embedded software transmitted and receiver, sense and send a signal to the causes the microprocessor to send a signal to the output system, at one end an infrared transmitter is fitted port of the interface unit in other to activate the dc which transmit a signal to another infrared receiver so motor to control the opening and closing of the gate. if this signal is interrupted by a vehicle, there is an input to the trigger circuit which is held HIGH. The A LOW will never activate the gate, false triggering is trigger circuit serves as an ADC (Analog-to Digital taken care of by circuitry. There is a DC power unit converter), which produces a HIGH when the beam is supplier which supplies the required voltage by the interrupted. The trigger circuitry sends a signal to the system and this is constant because the microcontroller interface unit, which is made up of Programmable is highly sensitive to voltage supplied into it.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2.1 System Design 2.4 The Trigger Circuitry

In an attempt to implement the proposed system, it was This is made up of two relays and two electric motor. It divided into two modules which are hardware and accepts the output from the sensor circuit. It is in such a software design considerations. way that only when there is an output from the sensing unit, the trigger circuitry goes HIGH else, it remain at 2.2 Hardware Design Considerations LOW

The hardware design of the proposed system consists of: the sensor unit, the trigger circuit, the microcontroller, the display unit, the gate control unit, the power supply unit. These parts are discussed as follows:

IN4001 2.3 The Sensor Unit

The sensor unit consists of infrared diodes which are of two types: A transmitter and a receiver. 33Ώ The infrared transmitter has the ability to transmit infrared beam but can only travel in a rectilinear + manner or a line of sight, which is received by the 10K infrared receiver at another end. TO PIC C945 M -

IN4001 Fig 2: Infrared transmitter

33Ώ

10K

TO PIC C945

Fig 4: Trigger Circuitry. Fig 3: Infrared receiver

The circuit has the ability to detect the passage of an 2.5 PIC16F84 Microcontrollers automobile through the entrance and the exit of the gate only if the infrared beam is interrupted from either side. The PIC16F84 belongs to a class of 8-bit Each pair of the sensor is separated by a reasonable microcontrollers of RISC architecture. The PIC chips distance such that the passage of a person or other have two separate 'data' busses, one for instructions and moving object cannot obstruct the sensor pair one for everything else. Instructions are essentially in separation. Also the height of the sensor is considered ROM and dedicates the microcontroller to doing one only the body of the vehicle can interrupt the light task, RAM is where variables are stored, there is very beam of the sensor and not the tires or its windows. little RAM, a few dozen bytes, and this is reserved for variables operated on by the program. There is also very little 'data' storage, again a few dozen bytes, and this is in EEPROM which is slow and clumsy to change. EEPROM is used to hold values to be remembered when the power is turned off

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Fig 5: PIC16F84 block diagram

2.6 PIC16F84 Central Processing Unit (CPU)

The PIC16F84 CPU is one of the most powerful 8-bit microprocessor in the electronics world today. It is a sophisticated, sequential, digital circuit that is designed to follow a sequence of instructions called a program. The program of instructions put in memory for the microprocessor to execute makes it so versatile and flexible in that its operation can be changed by simply changing the programs stored in memory (software) rather than rewire the electronics (hardware).It has a role of connective element between other blocks in the microcontroller. It coordinates the work of other blocks and executes the user program.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 6: PIC16F84 microcontroller outline

2.7 PIC16f84 PIN Description

PIC16F84 has a total of 18 pins. It is most frequently found in a DIP18 (Dual In Package) type of case but can also be found in Surface Mount Devices (SMD) case which is smaller from a DIP.

Figure 7: PIC16F84 PINS Layout

The Pins on PIC16F84 microcontroller have the Clock from the oscillator enters a microcontroller via following meaning: OSC1 pin where internal circuit of a microcontroller divides the clock into four even clocks Q1, Q2, Q3, and 2.8 PIC16F84 Clock/Instructions Q4 which do not overlap. These four clocks make up one Clock is microcontroller's main starter, and is obtained instruction cycle (also called machine cycle) during from an external component called an "oscillator”. The which one instruction is executed. Execution of small instruction set, (37 instructions), and the 14 bit size instruction starts by calling an instruction that is next in of instructions lead to a number of compromises. One string. Instruction is called from program memory on cannot have two registers specified in a single instruction. every Q1 and is written in instruction register on Q4. Each register takes 7 bits to specify its address, but one Decoding and execution of instruction are done between also have to specify the instruction number and what to the next Q1 and Q4 cycles. On the following diagram we do. By comparing a microcontroller with a time clock, can see the relationship between instruction cycle and the "clock" would then be a ticking sound we hear from clock of the oscillator (OSC1) as well as that of internal the time clock. In that case, oscillator could be compared clocks Q1-Q4. Program counter (PC) holds information to a spring that is wound so time clock can run. Also, about the address of the next instruction. force used to wind the time clock can be compared to an electrical supply.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Table 1: PIC16F84 PINOUT Description

Note I= Input O= Output I/O = Input/Output P = Power ST = Trigger input TTL = TTL input __ = Not used

Figure 8: Clock/Instruction Cycle

2.8 Supplying the microcontroller

For a proper function of any microcontroller, it is necessary to provide a stable source of supply, a sure reset when you turn it on and an oscillator. According to technical specifications by the manufacturer of PIC microcontroller, supply voltage should move between 2.0V to 6.0V in all versions. The simplest solution to the source of supply is using the voltage stabilizer LM7805 which gives stable +5V on its output. One such source is shown in the picture below.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 9: Supplying PIC16F84 microcontroller

In order to function properly, or in order to have stable 3. SOFTWARE DESIGN CONSIDERATIONS 5V at the output (pin 3), input voltage on pin 1 of LM7805 should be between 7V through 24V. Designing software for the automatic gate was not a Depending on current consumption of device we will very simple task. In the development cycle of a use the appropriate type of voltage stabilizer LM7805. Microcontroller-based system, decision is made on the There are several versions of LM7805. For current parts of the system to be realized in hardware and the consumption of up to 1A we should use the version in parts to be implemented in software. The software is TO-220 case with the capability of additional cooling. decomposed into modules so that each module can be If the total consumption is 50mA, we can use 78L05 individually tested as a unit and debugged before the (stabilizer version in small TO - 92 packaging for modules are integrated and tested as a software system current of up to 100mA). in order to ensure that the software design meets its specification [10]. The program for the system is 2.9 Control Unit written in Assembly Language for speed optimization. Assembly code represents halfway position between The gate control unit consists of these components: machine code and a high level language. The assembly NPN transistor, Diodes, Electric motors, and Relays. code is usually a mnemonic code derived from the The NPN transistors are arranged in such that a pair instruction itself, i.e. LDA is derived from Load the (NPN) controls the opening of the gate through the Accumulator. Assembly code is thus very easy to operation of relays and motors and the other pair remember and use when writing programs. When reverses the polarity of the motor by rotating it in the entering an assembly program into a microcontroller, opposite direction to close the gate. There is a time the assembly code must first be converted into machine interval of 5.0 Seconds between the openings of the code. For short programs, of a few lines, this is gate. The software in the PIC varies this time interval. relatively easy and usually requires that the The arrangement of the diodes serves to protect the Programmer has next to him or her, a table which transistors from reverse biased polarity and the resistors contains the assembly mnemonics and the equivalent serve to improve switching time. The motor is used to machine code. control the opening and closing of gate. The electric (DC) motor used is the one that have the ability to This technique is known as Hand Assembly and is rotate in both directions simply by reversing the limited to programs of about one hundred lines or less. polarity. In the case of longer programs, a separate program called an assembler program is used to convert the assembly code into machine code, which is placed directly into the microcontroller memory. The program modules are segmented into: the main program, the sensor subroutine, delay subroutine, and the output (Gate Control) subroutine.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

START

Setup port and pin Directions

Reset all Registers and Clear Ports

Get Status of Sensor Code

Is there Code for Entrance? NO S E Y

Enable the Gate to Open

C A B D

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

C A B D

Close the Gate After 5 seconds.

Update Display Using a Shift Register

Test if Reset Button is Pressed

Reset Yes Button is Pressed

Figure 10: Flowchart of microcontroller based automatic gate

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4. SYSTEM CONSTRUCTION, TESTING AND The test instrument used for examining logic signal, RESULT testing and troubleshooting applications in the course of the project were: digital multimeter, logic probe, and The stages involved in the construction of the proposed oscilloscope. Testing involve troubleshooting the system are: design validation, bread board and vero hardware system to detect, isolate and correct internal board implementations, testing, packaging and or external fault such as malfunction in the internal modeling. circuitry, input or output shorted to ground or Vcc input or output open circuited, short between two pins broken 4.1 Design Validation wire, poor of dry connection, bent or broken pins, or an IC and faulty ICs socket. The hardware system was The best workable circuit was devised taking into properly tested because the software cannot work when consideration some parameters such as signal levels the hardware is not functioning properly. between components, compatibility of signals and components, cost and availability of components. The The testing of the entire circuit was carried out in program to direct the operation of the gate was written stages in the Assembly Language and electronically written 1. Each of the components was first tested using into the PIC (16F84A). A very important advantage to the multimeter in order to check for their design validation is the use of software packages such state of performance and accurate values. as ORCAD PSPICE to simulate the design before 2. In the connection of each component on the implementation. Vero board was then tested. This was done in other to carry out the continuity, which is 4.2 Breadboard Implementation meant for proper connection of the circuit and to detect any wrong connection. The circuit design was implemented on the bread board 3. The sensor unit circuitry was tested to after validating it on circuit simulator. During this ascertain the degree of sensitivity. A small stage, various parameters like voltage drops, input prototype car (object was placed between the impedance, base current, pulse width were measured in two pairs of the infrared diodes to obstruct order to ensure good result. light rays. The voltage levels at the output The circuit design was tested on the board and found to were observed with the aid to a digital be working properly before soldering. multimeter. The result is shown in the table 2.

4.3 Vero Board Implementation Table 2: Voltage levels of the sensor unit

After proper verification on the breadboard, the design Test Result was transferred to a Vero board for permanent Without object 0.65V construction. The various module of the design were With object 5.00V soldered and arranged on the Vero board such that each Object removed 0.70V module can be easily identified. Before proper soldering, component layout plan was Before the result was obtained, the variable drawn paying particular attention to minimizing the resistor was adjusted to obtain the output voltages. distances involve between point to be connected and 4. The output of the trigger circuitry was tested the prevention of the overcrowding. All other by connecting LED across to check if it lit or components were then connected up to implement the not a lit indicate the presence of a low logic circuit. as shown in table 3 below.

4.4 Testing and Result Table 3: Trigger circuit logic level

With the advent of digital systems and in particular Sensor voltage levels Output logic. microprocessor based ones, new tools and techniques V01 V02 F have been developed for testing and to carry out troubleshooting. Vast amount of digital information flow, for example, over the busses of a microcomputer 0.70 0.70 0 system and even a single faulty chip or a single 0.70 5.00 0 incorrect bit can lead to a total system malfunction. It is 5.00 0.70 0 therefore of paramount importance to establish a highly 5.00 5.00 1 efficient testing techniques in other to minimize cost.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

5. Also, the gate control circuit was tested by 6. RECOMMENDATIONS applying logic 1 or 0 to point A and B of the circuit. When logic 1 is applied to point A, The following suggestions should be considered in The motor rotate in a clockwise direction carrying out further work on this study: while logic 1 at point B changes the direction of the motor. Logic 0 at both points will A more effective and sensitive sensor is never activate the motor. The result is shown recommended for better performance. For in table 4. example a sensor such as RADAR sensor that could detect contraband goods in vehicles. Table 4: Gate Control circuit truth table The achievement of a full automation, a real time system may be employed and a A B Motor biometric scanner that will provide a proper direction monitoring and security purposes. This shall 0 0 Inactive be helpful in tracking the identity of the 0 1 Anticlockwise vehicle before the system is activated. 1 0 Clockwise 1 1 Inactive REFERENCES

6. After the proper testing of the peripherals and [1] Rafiquzzaman, M (2011) 'Microcontroller found to be working perfectly, the entire Theory and Applications with the PIC 18F" circuit was tested. Series of programs [2] Di Jasio, L (2008) "Programming 32-Bit {software} were written and tested before the Microcontrollers in C: Exploring the PIC 32 working program was fully achieved. The (Embedded Technology)" circuit worked perfectly as designed. The [3] Wilmshurst, T (2009) "DESIGNING display unit was also observed during the Embedded Systems with PIC testing. Microcontrollers: Principles and Applications" Packaging [4] Reese, R.B., Bruce, Bruce, J. W. And Jone, B. A. (2008) "Microcontrollers: From After proper testing was conducted, the packaging of Assembly Language to PC using the PIC the design into a model and casing was considered. The Family" connecting wires were properly connected and well [5] Morton, J (2005) "The PIC Microcontroller: insulated, also the wires were well packed and bounded Your Personal Introductory Course" together. [6] Huang, H and Chartrand, L (2004) "PIC Microcontroller" An Introductory to Software 5. CONCLUSION and Hardware Interfacing" [7] Sandhu, H (2008) "Making PIC The use of microcontroller system has been achieved in Microcontroller Instruments and Controllers" the design and implementation of this project. This [8] Bates, M. P (2011) "PIC Microcontrollers: project can be easily tailored to any electric gate and all An Introduction to Microelectronics" kinds and all forms of control, which has the use of [9] Van Dam, B (2008) "PIC Microcontrollers: sensors. For an effective design of this kind of system it 50 Projects for Beginners and Experts" [10] is imperative to have a good grasp of the basic sensor Prof. A.O Odinma, Software Engineering characteristic, microcontroller characteristic and Lecture Notes assembly language principles. The infrared photodiode [11] Prof. A.O Odinma and Engr. O. Shoewu, which was used as a sensor serves as a transducer for Computer Aided Application Lecture Notes vehicle detection while the programming language is [12] Engr. Lawrence Oborkhale, Microprocessor fundamental to software design based on the system and Applications Lecture Notes requirement, specification and operation of the system. [13] Engr. Balogun, Semiconductor Devices The automatic gate designed can be used in companies, Lecture Notes public car park, domestic parking lot and automobile [14] http://www.privatedoor.com terminal, where little or no form of security is required. [15] http://www.access-automation.co.uk [16] http://www.gateautomation.com [17] http://www.twystedpair.com [18] http://www.howstuffswork.com

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

APPENDIX

12V 12V 12V 12V 2.2m 10kΩ 10kΩ 2 10kΩ 3 0.1m F 10kΩ 4.7mF 220 Ω 2.2mΩ 10kΩ

47kΩ C945 10kΩ 22kΩ 10 KΩ

10kΩ +5V OPEN SWITCH 100pF 12v 12V

4 PIC 1 2 IN4001 5 16F84 + 3 M 33Ω - 10kΩ LOCK SWITCH C945

10mF

12V 12V

4 470Ω 470Ω 1 2 IN4001 5 3

5V 5V 33Ω -- 10kΩ C945

The Main Circuit Diagram of the Proposed System

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Understanding the Potential of Data Mining in Botswana

George Anderson, Audrey N. Masizana-Katongo, and Dimane Mpoeleng Department of Computer Science University of Botswana P/BAG UB 00704 Gaborone, Botswana {andersong,masizana,mpoeleng}@mopipi.ub.bw

ABSTRACT Botswana is a rapidly developing country, with many new organizations establishing presence every year. With this growth in number and size of organizations comes the generations of large amounts of organizational data. These organisations are faced with the challenge to analyze this data effectively and efficiently in order to gain important strategic advantage and competitive edge over their rivals. However, the huge amount of data available has surpassed human cognitive and analytical abilities. Hence strategists and decision makers require effective and efficient data analysis software in order to undertake thorough analysis on the data to reveal knowledge structures that can guide businesses and decision-making processes. On one hand, IT infrastructure is also growing very fast, coupled with advances in both computer hardware and software including scalable algorithms and data extraction tools. One such technology is data mining which is, as defined by Giudici, a process of selection, exploration and modeling of large databases in order to discover models and patterns that are unknown a priori. Data mining tools are used to explore summaries, comparison, analysis, forecast, estimate of the data. In this paper we explore how organizations in Botswana can take advantage of data mining if they are not already doing so. We envisage how they could realise potential usefulness of data mining in transforming raw data into valuable information. We highlight key data mining technologies that could be useful and use some of these technologies to process organizational data in an experiment using data mining software. Through the experiment we demonstrate the value of this critical technology. We also present case studies of companies in other countries that have benefitted from data mining.

Keywords- Data Mining, ICT for Development, Botswana

African Journal of Computing & ICT Reference Format: George Anderson, Audrey N. Masizana-Katongo & Dimane Mpoeleng (2013 Understanding the Potential of Data Mining in Botswana. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 33-42.

1. INTRODUCTION

Information has become a key resource in today’s A lot of research has focused on developing these business world and an ability to effectively manipulate technologies that can assist in analysing data for it has become vitally important to most organisations. improved business and decision-making processes One Organizations and individuals having access to the such technology is commonly referred to as data right information at the right moment, have greater mining, described as “the process of selection, chances of being successful in the globalized and cut- exploration, and modeling of large quantities of data to throat competition. At the heart of this process is data discover regularities or relations that are at first that is only useful if it reveals valuable knowledge unknown with the aim of obtaining clear and useful hidden within it. Humans have been "manually" results for the owner of the database” Giudici [1]. This extracting patterns from data for centuries however technology has been identified as one that can be used modern organisational data has increasingly grown in to extract diamonds of knowledge from large datasets size and complexity. The huge size of these data and predict outcomes of future situations thus helping sources make it impossible for a human analyst to to optimise decision processes. It involves the use of discover valuable information hidden in these data sophisticated data analysis tools to discover previously repositories. These datasets require sophisticated unknown, valid patterns and relationships in large data processes that should filter and convert the data into sets [2]. Also, in most standard database operations, required useful information. nearly all of the results presented to the user are something that they knew existed in the database already.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Data mining, on the other hand, extracts information Various data mining software tools are available. Some from a database that the user did not know existed [3] are built into commercial database managements and in contrast, utilizes a discovery approach which systems. These tools provide a variety of data mining sifts through layers of seemingly unrelated data for algorithms. For example Microsoft’s SQL Server has meaningful relationships. Data mining also differs from the following built in algorithms [10]: traditional statistical techniques in the way the data is mined and usually works on messier real world data.  Microsoft Naïve Bayes. This algorithm calculates correlations between a variable of interest and all Although data mining is a relatively a new term, the other variables. This could then be used to technology is not. As data sets have grown in size and determine which category the variable belongs in. complexity, companies have used powerful computers It could also be used to determine factors that technology to sift through volumes of data for years. differentiate categories. Several recent trends have increased the interest in data  Microsoft Decision Trees. This algorithm is mining, mainly due to the declining cost of data storage mainly used for classification. Given an object, and computational power and the development of questions are asked of it in a hierarchy. Each robust and efficient machine-learning algorithm [4]. alternative answer (e.g. Yes or No) directs the These algorithms empower the data mining process to object to the next question below in the Decision be able to answer difficult questions or problems that Tree. At the end, the object ends up in a low level would be too time-consuming and/or complex to of the tree that identifies the category it belongs to. resolve using traditional methods. It is often considered  Microsoft Time Series. This algorithm is used to be "a blend of statistics, AI (artificial intelligence), mainly to track trends over time and forecast and data base research" [5]. values (e.g. sales figures, so that appropriate inventory can be maintained).  Microsoft Clustering. This algorithm groups Data mining is becoming increasingly common objects (e.g. customers) according to values of worldwide in both the public and private sectors certain attributes. Grouping is easy for humans to Industries such as banking, insurance, medicine, and do when each object has just a few attributes but retailing commonly use data mining to reduce costs, becomes complex with dozens or even hundreds enhance research, and increase sale [6]. Data mining of attributes. applications can use a variety of parameters to examine the data. The patterns and hypothesis are automatically  Microsoft Sequence Clustering. This algorithm extracted from data rather than being formulated by a finds sequential patterns of certain events, e.g. user as it is done in traditional modeling approaches, Web page hits, and groups similar patterns e.g., statistical or mathematical programming together, e.g. all visitors to a Web site exhibiting modelling [7]. Data can be mined to identify similar browsing behaviour will form a separate associations where the objective is to determine which group. variables go together, sequences of anticipated patterns  Microsoft Association Rules. This algorithm is and trends in the data, classes of entities represented in usually used for Market Basket Analysis e.g. to data, clusters of data with objects in the same cluster find which items are usually bought together with similar and objects belonging to different clusters which other items. In other words, which objects dissimilar, correlations between variables of interest usually go with which other objects. and forecasting regarding the future.  Microsoft Neural Network. This algorithm is usually used for classification and regression These processes are conducted using data mining tasks. It can find nonlinear relationships among algorithms such as clustering, sequencing, association attributes. It is a more complex algorithm than algorithms. Other includes neural networks, described Decision Trees or Naïve Bayes and interpreting as “a model of reasoning based on the human brain" results can be difficult, but it is very powerful [8]. This process is capable of predicting new Association Rules, Decision Rules, Sequence observations from other observations and can be Clustering, and Naïve Bayes. applied in classification and prediction processes. Researchers also employ decision trees described as a Oracle also offers data mining techniques packaged in tree from a data set to classify objects with unknown its Oracle Data Mining (ODM) software [11], with decision [7]. Data mining can also apply Genetic models included in SQL queries and embedded in algorithms which are optimization techniques based on applications to offer improved business intelligence. the concepts of evolution such as genetic combination, Hamm [12] provides an easy to read, step-by-step, mutation, and natural selection in a design [9]. The practical guide for learning about data mining using choice and application of technologies would depend Oracle Data Mining. Data mining also features in open largely on the size of the database and the complexity source software for example, Hall [13] and Keel [14]. of the queries on the data required from the databases.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The right algorithm (or algorithms) has to be used for a Rajanish Dass [20] and Madan Lal Bhasin [4] discuss particular problem. the potential of data mining in bank processes such as Customer Acquisition and Retention, Fraud Detection, This paper looks at the potential usefulness of data Risk Management, Marketing and Trading. Bank of mining in Botswana organisations and presents an America's mortgage division has used data mining on illustrative application focused on the data mining customer behaviour data to estimate bad loans, so that methodology and its tools are also discussed and the credit risk managers can allocate optimal loan loss data mining literature summarised. reserves which affects profitability directly [21]. Data mining technology also has the ability to profile 2. APPLICATIONS OF DATA MINING common usage scenarios and detect new or different patterns which can be used for fraud detection, In literature, data mining applications can be found in prevention or general investigations. Hsinchun Chen et surveillance, traffic analysis, government operations, al [22] review data mining techniques applied in the bioinformatics, genetics, medicine and education. In context of law enforcement and intelligence analysis, this section we give examples of application in and present four case studies done in the ongoing healthcare, finance and retail, and some educational COPLINK project [23]. The paper presents various processes that could be relevant to the environment in crime data mining approaches and techniques, for Botswana. example entity extraction to extract personal properties from police narrative reports, clustering techniques to 2.1 Healthcare associate different objects (such as persons, organizations, vehicles) in crime records, classification Major initiatives to improve the quality and success to detect email spamming and find authors who send rate of healthcare data delivery are emerging all over out unsolicited emails etc. the world. Data mining technologies have been employed in health care to enhance these processes. Breault et al [15] examine data mining applied to a 2.3 Academic Institutions diabetic data warehouse, showing a method of applying The advanced networking infrastructure, connectivity data mining techniques, and presenting some of the to digital educational environments has made the data issues, analysis problems, and results. The diabetic collection, transfer, and dissemination of academics data warehouse is from a large integrated health care related information more accessible than at any other system in the New Orleans area with 30,383 diabetic point in history. This has enabled new technologies patients. such as data mining to be applied in academic environments in order to improve decision making In epidemiological research, the STATISTICA [16] processes. Academic institutions regularly generate data mining software system is used to identify factors huge data on students, courses, faculty, and staff that increasing the risk of diseases occurring. The software includes managerial systems, organizational personnel, is also used for planning and data analysis for clinical and lectures details and so on [24]. research associated with the development of new treatment methods, analysis of survival rates and This useful data serves as a strategic input to any factors influencing prognosis. Cios et al [17] address academic institution for improving the quality of the special features of data mining with medical data. education process. Merceron et al [25] show how using They address ethical and legal aspects of medical data data mining algorithms can help discover pedagogically mining including data ownership, fear of lawsuits, relevant knowledge contained in databases obtained expected benefits, and special administrative issues. from web-based educational systems. They use Data mining tools has also been employed to mine association rules to find mistakes often occurring large collections of electronic health records for together while students solve exercises, decision trees temporal patterns associating drug prescriptions to to try and predict exam marks and clustering and medical diagnoses [18]. cluster visualisation to identify a particular behaviour among failing students. They argue that their findings 2.2 Finance and Retail Institutions can be used both to help teachers with managing their class, understand their students’ learning and reflect on Currently, huge electronic data repositories are being their teaching and to support learner reflection and maintained by financial and retail institutions across the provide proactive feedback to learners. globe. This has been due to ‘electronic’ business which has made acquisition of transactional data to become easier and to grow exponentially. Data mining tools have now been used to mine the complex data for strategic information. In particular, data mining is widely used in marketing, risk management and fraud control [19].

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Erdogan et al [26] apply data mining to study the Data mining technology has the ability to turn data into relationship between students university entrance actionable information that companies can use to examination results and their success using cluster transform the way of interaction with service recipients. analysis and k-means algorithm techniques. Behrouz It would provide the ability to proactively make Minaei-Bidgoli [27] apply data mining approach in changes upon future needs. The country’s growing ICT classifying students in order to predict their final grades infrastructure is also enabling organisations to explore based on features extracted from logged data in an new technologies. Published research on data mining in education web-based system. They apply classification the Botswana has, to date, been limited. Most techniques to form classes of performances and genetic Companies have acknowledged the need to obtain algorithms to improve the results obtained and value from the data stored in their data warehouse. conclude that data mining efforts can be useful in There have been various studies related to data mining predicting student outcomes. A study by Fadzilah et al activities in Botswana. Mogotsi [32] studied the use of [28] presents the results of applying data mining to text mining on stories published in the Daily News, a enrolment data at Sebha University in Libya in order to Botswana government newspaper. guide the enrolment processes and improve on conventional processes. The study tries to find natural groupings of news stories. This could be used to find out what sorts of Cluster analysis was performed to group the data into stories were deemed interesting and what sorts of clusters based on its similarities. They also apply stories were reported using government funds. Such an Neural Network, Logistic regression and the Decision approach could also be used for Competitive Tree for predictive analysis. Teh et al [29] use data Intelligence (CI), where companies try to learn about mining to audit usage of stationeries in the Faculty of their competitors’ activities. Golosinski et al [33] Computer Science and Information Technology, applied data mining to data generated by sensors in a University of Malaya in Malaysia. They collect data truck used for mining activities in Botswana. Their from the log file for user and based on this they are able analysis revealed such information as relationships to discover the frequently access attributes by using between various operating conditions and performance association rule algorithms. parameters, with the ability to predict certain parameters. 3. POTENTIAL OF DATA MINING IN BOTSWANA A paper by Hart [34], describes three largely qualitative studies, spread over a five year period (1997-2002), The Government of Botswana has demonstrated a very into the current practice of data mining in several large high level of political commitment in ICT development South African organisations. He concludes that thorough Maitlamo [30], its National Information published research in data mining in South Africa is Communications Technology (ICT) strategy. This just as limited although progress is revealed to have Policy provides Botswana with a clear and compelling been made over this period. He adds that by 2002 a roadmap that will drive social, economic, cultural and major shift had taken place, with interviewed political transformation through the effective use of organisations all having gained experience with data ICT. It strives to transform the nation into a digitally mining. His study revealed that companies currently competitive technology aware and globally competitive performing data mining are successfully adopting wide ICT nation. To enter the globalized world it is range of techniques, ranging from conventional to necessary for Botswana to become part of the unique, tailored uses. information society. As illustrated from case studies and experience in The advent of ICT in the recent years has presented an South Africa, there is potential for data mining in opportunity for the way organisations leverage and Botswana organisations. Data mining can be applied value their information assets. The timely adoption of in many environments and the following are just technology and advances in technology are of interest examples of potential application. to all organisations in Botswana since information technology usage fundamentally alters the domains within which it is implemented [31]. The Adoption and adaptation to new technologies such as data mining, we believe, would benefit most organisations in the country to enhance the quality of information available for decision making processes.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

3.1 In Government 3.3 In Academic Processes Government Departments could use data mining to Data mining can be used for educational processes study classes, patterns, causalities, sequences, episodes similar to [24-30]. In general, it can be used in and correlations in that large data sets already collected analyzing data for patterns, such as students’ over time. The utilisation of data mining tools could interactions and evaluations, admissions processes and influence change in practices and impact on future procedures, courses and subjects management to assist decision making processes and help in formulating the administration, faculty, and staff. Classification and more effective strategies and policies for citizen Prediction models can be used to describe data classes facilitation. like student performance levels. Clustering can be used to build hierarchy of classes that group similar students These applications may include budgetary plans, based on educational background, age, areas of interest financial markets, import and export analysis, resource and specialization and so on. Decision trees, Bayesian allocations, HR activities, crime patterns, demographic and Regression models can be applied for prediction of change predictions, immigration patterns, tourist student’s choice of specialization, student enrolment records, health, and educational trends. Health processes, student retention etc. Associations could be Services could employ data mining techniques to study used to track students activities related to attendances, patient trends e.g. outpatient revisits, predict the specializations and courses effectiveness of procedures or medicines and help guide research on new treatments for diseases. 4. APPLICATION OF DATA MINING IN A PUBLIC UNIVERSITY IN BOTSWANA 3.2 In Commercial Industry Commercial Industries such as banks, retailers, and To demonstrate the usefulness of data mining in service providers can study customers’ past purchasing Botswana, we carried out a study of its application in a histories for better customer retention and relationship. public university, the University of Botswana (UB). Through data mining, they could know what kinds of The University of Botswana is undergoing promotions and incentives to target to which type of organizational restructuring [35]. A new level is to be customers, etc. This can also help in adjusting introduced between academic departments and the relationship with these customers so that the loss in faculty level: the school. Therefore departments will be future is kept to its minimum. They can also improve grouped together to form schools, which will lie under on better targeting and acquiring of new customers the governance of a faculty. Various constitutions of using descriptive and predictive data mining techniques schools have been proposed. We propose an automated such as clustering, classification and sequencing approach to the formation of schools, based on various algorithms. attributes of academic departments.

Data mining techniques can also improve marketing by 4.1 University of Botswana Organizational Data modelling customers’ past and future demand in order to design calculated and informed marketing strategies. For our experiment we generated data about imaginary They can also be applied for risk management to departments in UB. For 10 imaginary departments provide extensive scenario analysis where capital can (Department A – Department J), we broke up their be allocated to business activities to maximise profit activities into the following categories: Undergraduate and minimise risk. For example banks can use decision Projects, Undergraduate Courses, Graduate Courses, trees to score attributes of each customer account for MSc Projects, PhD Projects, Funded Projects, Non- credit risk assessment or credit scoring. In fraud Funded Projects, Conference Publications, Book detection, organisations may use data mining to Publications, Journal Publications, Workshops understand which products or services may be Organized, Seminars Organized, Service to UB, vulnerable for theft and deal with these situations Service to the Profession, and Service to the accordingly. Data mining can also be helpful to human- Community. Each attribute (category) has an integer resources departments in identifying the characteristics value attached to it. This value specifies the area of of their most successful employees which can help HR academic knowledge that category was focused on. For focus recruiting efforts accordingly. example, a department could have its PhD projects focused on area 15. We could have many possible areas of academic knowledge, 1,2,3,…, etc, going up to 60, or even more if desired. This data was generated randomly using a normal distribution with different means for different departments. As a result, departments could have academic areas overlapping.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4.2 Data Mining Analysis of the University of 5.1 Bank Data Botswana Organizational Data For our experiment we generated data about imaginary To analyze our data, we used a classification algorithm historic loan applications by imaginary customers (10 called Sample K Means. This algorithm analyzes the of them). Historic means the applications happened in data and groups departments together that have similar the past. Each application has the following attributes: academic areas of emphasis in their activities in Customer, Gross Salary, Net Pay, Loan Amount, Loan clusters. We used a free open-source Data Mining tool Period, Age, Affiliation, Education, Marital Status, called Weka [13]. A graphical result of the analysis is Children, Num Cars, and Loan Granted. The target displayed in Fig. 1 in the Appendix. variable, Loan Granted, is either Y or N (Yes or No). Each attribute has a set of possible values, e.g. for Num 4.3 Discussion of Findings Cars, the values are 0, 1, 2+ (no cars, one car, two or more cars). Fig. 1 shows the results of the data mining analysis.

Two clusters have been identified: Cluster 0 and 5.2 Data Mining Analysis Cluster 1. Departments A, B, C and D belong to Cluster 0, while E, F, G, H, I and J belong to Cluster 1. The We selected a Decision Tree algorithm available in clusters are formed based on the similarities of Weka called Random Tree [13]. Fig. 2 in the Appendix attributes between the various departments. Since the shows what the tree created looks like. constituent departments are similar, each cluster could form a school. Note that randomly generated data was 5.3 Discussion of Findings used, influencing the results that were obtained. However, this experiment validates the usefulness of Not all attributes feature in the tree displayed in Fig. 2. clustering within a university like UB. In particular, to This is because based on the historic data provided, guide the formation of schools. known as the training data, some attributes have negligible influence on the decision to approve a loan or not. From the figure, the most important input is the 5. APPLICATION OF DATA MINING IN THE education level of the applicant. BANKING SECTOR IN BOTSWANA For example, if the applicant has a primary school certificate, we then consider the amount of the loan To demonstrator another possible application of data application. If it is 10K-49K, we approve the loan. The mining in Botswana, we carried out a study on the generated tree will be used by a computer program to process of loan approval by banks, using information help bank staff make decisions about loans. Note that commonly sought by banks in this process. the historic data was randomly generated using a uniform distribution, so some decisions to award a loan Over a period, a bank collects data from its clients, or not may not make perfect sense. However, what is analyses it, and makes intuitive decisions based on the important is the model that is built: the decision tree. If collected client’s data and experience of the loan we have real data, we could simply run the algorithm approver. However this process is not efficient in that it on the real data and create a decision tree that actually may miss client’s attributes and behaviour patterns that makes real-life decisions that make sense even to can only be best extracted by deploying data mining experts. techniques. A data mining process would reveal those patterns in the banks historical data to assist bank manager in making valuable decisions. Decisions that are value added are those that positively affect the bottom line of the business, in this case to make profit thereby stay competitive against business rivals. [36, 37, 38]

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

6 CONCLUSION REFERENCES

We conclude that the computational complexity and [1] Giudici, P. Applied Data Mining: Statistical robustness of knowledge extraction from large data sets Methods for Business and Industry. Chichester: and decision making can be enhanced by data mining John Wiley & Sons Ltd, 2003. as observed from the literature review. We have also [2] Adriaans, P., and Zantinge, D., Data Mining, demonstrated through examples of case scenarios in New York: Addison Wesley, 1996. academic processes and banking in Botswana. We [3] Thearling K, Understanding Data Mining: It's encourage Botswana organisations to explore data All in the Interaction, Journal for Data-Intensive mining technologies and tap into its potential. Decision Support, 1(10), December 1997. [4] Lal Bhasin, M., Data Mining: A Competitive Our next research objective is to explore this potential. Tool in the Banking and Retail Industries, We want to begin by performing an analysis of the Banking and Finance, The Chartered specific opportunities for data mining in Botswana, Accountant, page 590-594, October 2006. identifying specific data stores where the impact of data [5] Pregibon, D. Data mining, Statistical Computing mining will be greatest, and analyze the data using data and Graphics, 7(8). 1997. mining technologies. We will then measure the value of [6] Seifert, J.W., Data Mining: An Overview, the analysis. Congressional Research Service Report for Congress, The Library of Congress, December 2004. [7] Kusiak A, Decomposition in Data Mining: An Industrial Case Study, IEEE Transactions on Electronics packaging manufacturing, 23(4), October 2000. [8] Negnevitsky M, "Artificial Intelligence, A Guide to Intelligent Systems", England: Pearson Education Limited, 2002. [9] Goldberg D, Genetic Algorithms in Search, Optimization and Machine Learning, Addison- Wesley Professional, 1 edition (January 11, 1989) [10] Tang, Z., and MacLennan, J. (2005). Data Mining with SQL Server 2005. Indianapolis: Wiley Publishing, Inc, 2005. [11] Oracle Data Mining Software, http://www.oracle.com/technology/products/bi/o dm/index.html, retrieved July 2009. [12] Hamm C, Oracle Data Mining: Gold from your Warehouse, Rampant Techpress, Jul 2007. [13] Hall, M., Data Mining with Open Source Machine Learning Software in Java: http://www.cs.waikato.ac.nz/ml/weka/, Retrieved June 23, 2009 [14] KEEL, Knowledge Extraction based on Evolutionary Learning: http://www.keel.es/, Retrieved June 10, 2009 [15] Breault J, Goodall C, Fos P, Data mining a diabetic data warehouse, Artificial Intelligence in Medicine, 26(1-2), 37-54. [16] STATISCA data miner: http://www.statsoft.com/products/statistica-data- miner/, retrieved, July 2009 [17] Cios K., Moore, W., Uniqueness of medical data mining, Artificial Intelligence in Medicine, 26(1- 2), pp. 1 – 24. [18] Data Mining Definition: Wikipedia, http://en.wikipedia.org/wiki/Data_mining, July, 2009

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

[19] Kuykendall, L. The data-mining toolbox. Credit [28] Siraj, F., Abdoulha, M.A., Uncovering Hidden Card Management, 12(6): 30-40. 1999 Information Within University's Student [20] Dass, R., Data Mining In Banking And Finance: Enrollment Data Using Data Mining, A Note For Bankers, Indian Institute of Proceedings of the Third Asia International Management Ahmadabad Conference on Modelling & Simulation, pp. 413- http://www.iimahd.ernet.in/publications/data/No 418, 2009. te%20on%20Data%20Mining%20&%20BI%20i [29] Wah, T.Y., Nor, M.K.M., Bakar, Z.A. and Peck, n%20Banking%20Sector.pdf , retrieved, June L.S., Data Mining in Computer Auditing, 2009. Informing Science InSITE, pp1570, June 2002. [21] Fabris, P., Advanced Navigation, CIO, 11(15): [30] Botswana National ICT Policy: Mailtamo, pp. 50-55, 1998. http://www.maitlamo.gov.bw/, , retrieved, July [22] Chen, H., Chung, W., Qin, Y., Chau, M., Xu, 2009 J.J., Wang, G., Zheng R., and Atabakhsh, H., [31] Anderson K.V, Danziger J.N, The Impacts of Crime Data Mining: An Overview and Case, Information Technology on Public Proceedings of the 2003 annual national Administration: An Analysis of Empirical conference on Digital government research Research from the "Golden Age" of 2003, Boston, MA, May 18 - 21, 2003. Transformation, International Journal of Public [23] Hauck, R.V., Atabakhsh, H., Ongvasith, P., Administration, Vol. 25, 2002. Gupta, H., and Chen, H, Using Coplink To [32] Mogotsi, I. C., News analysis through text Analyze mining: a case study. VINE , 37 (4), pp. 516- Criminal-Justice Data. IEEE Computer, 35(3), 531, 2007. pp. 30-37, 2002. [33] Golosinski, T. S., Hu, H., and Elias, R., Data [24] Ranjan, J., and Malik, K., Effective educational mining VIMS data for information on truck process: A data-mining approach, Journal of condition. In: Computer Applications in the information and knowledge management systems Minerals Industries, Editors: Xie, Wang, & 37(4), pp. 502-515, 2007. Jiang, pp. 397-402, 2001. [25] Merceron, A., and Yacef , K. Educational Data [34] Hart M, Progress of Organisational data mining Mining: a Case Study. Proceedings of in South Africa , ARIMA/SACJ, No. 36., 2006 ArtificialIntelligence in Education (AIED2005), [35] University of Botswana, Consultation Paper: Amsterdam, The Netherlands, IOS Press, 2005 Revision of Academic Organizational Structure, [26] Erdoğan Ş.Z, and Timor M, A data mining 2009. application in a student database, Journal of [36] Bhasin, M.L, Data Mining: Acompetitive Tool in Aeronautics and Space Technologies, 2(2), pp. Banking and Retail Industries, The Chattered 53-57, July 2005. Accountant, pp588, October 2006. [27] Minaei-Bidgoli, B.I., Kashy, D.A.,, Kortemeyer, [37] Nuyagas, W., Srivihok, A., and Kitisin, S, G., and Punch, W.F., Predicting Student Clustering e-Banking Customer using Data Performance: An Application Of Data Mining Mining and Marketing Segmentation, ECTI Methods With An Educational Web-Based Transactions on Computer and Information System, Proceedings of the 33rd ASEEI/IEEE Technology Vol.2, No.1, pp63-69, May 2006. Frontiers in Education Conference, November [38] Scott, R.I., Svinterikou, S., Tjorjis, C., and 2003, Boulder, CO Keane, J.A., Experiences of using Data Mining in a Banking Application. Work at Department of Computation, UMIST, supported by ESPRIT HPCN Project no. 22693

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Author’s Brief

Mr. George Anderson is on Dr. Dimane Mpoeleng is a the faculty at the Department of lecturer at the University of Computer Science, University Botswana, Computer Science of Botswana, Gaborone, Department. He has been Botswana. His research has involved in a number of research spanned the areas of Operating activities including being the Systems, Machine Learning, principal Investigator for a Optimization, Information Mobile Telephony and Health Retrieval, Health Care Information Systems and Care Research Project that was Computing Education. He teaches courses in Operating sponsored by Microsoft, doing research on Distributed Systems and Artificial Intelligence, in addition to other Systems and Fault tolerance protocols, writing several areas, and supervises MSc students. George Anderson papers, journal articles, book chapter, and carrying out is due graduate with a PhD from the University of MSc and PhD student supervision. He is also the author Johannesburg, South Africa, in 2013. He can be of three computer awareness books published by reached through email at Longman/Pearson. Dr. Mpoeleng has attained his PhD [email protected]. from the University of Newcastle upon Tyne, England, UK He can be reached through email at [email protected]. Dr Audrey Masizana-Katongo is a senior lecturer at the University of Botswana, Computer Science Department. She holds a PhD in computer science from the UMIST in the UK (2004), and teaches both undergraduate and graduate courses in areas of Decision Support Systems, Expert Systems, Data Warehousing and Web Engineering. Her research interests are mainly in the area of Decision and Intelligent systems, particularly application of mathematical and computing techniques to decision problems. She has been involved in a number of research activities including Mobile Telephony and Health Care Research Project that was sponsored by Microsoft, Indigenous Knowledge Systems. She has written several papers, journal articles, book chapters, and also carries out MSc and PhD students’ project supervision. She can be reached through email at [email protected]

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Appendix

Figure 1: Weka cluster visualization. X axis shows departments, Y axis show academic area values for PhD projects in each department

Figure 2: Weka classifier tree visualizer. This shows the decision tree. First questions: What is the loan applicant’s eduction level? If Primary, how much is s/he asking for? If 10K-49K, approve the loan, etc.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

A Framework For Knowledge-Based Ontology Model In African Traditional Medicine

Omotosho, L.O. Department of Computer Science Afe Babalola University Ado-Ekiti, Nigeria.

Odejobi, O.A Department of Computer Science and Engineering Obafemi Awolowo University Ile-Ife, Nigeria.

Akanbi C.O. Department of Information and Communication Technology Osun State University Osogbo, Nigeria.

ABSTRACT Africa is a continent of diversified trado-medical practices, heritage and culture. However, there are increasing efforts directed at providing formal framework to describe the use of terms and relations in ontology for knowledge representation. In this paper, we describe a new approach to develop a formal description in ontology model in order to document, reuse and promote knowledge sharing among practitioners of Yoruba in south western part of Nigeria. Effective knowledge representation requires standardization of ontology development model. The main objective of this paper is to build an ontology model by grounding the used formalism in logics to represent formally, all the semantics underlying the concept behind this knowledge used in the domain of ATM. By reviewing the architecture of the existing works, we redefine some concept to formally and clearly describe the roles and terms involved. We used Logic-based Inference (Description Logic), facets coupled with DL rule of inference and the ontology knowledge based editor Prot´eg´e to formalize our approach.

Keywords- Ontology model; Knowledge representation;Ontology Web Language; hidden sematics;Description logics; African Traditional Medicine

African Journal of Computing & ICT Reference Format: Omotosho, L.O., Odejobi, O.A. & Akanbi C.O. (2013). A Framework For Knowledge-Based Ontology Model In African Traditional Medicine. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 43-48

1. INTRODUCTION

For thousands of years, African people have practised However, following the trend, this knowledge tends to various forms of heal- ing and medicine that involves decrease when practitioner of ATM decides not to both natural or supernatural explanations and remedies. explicitly share the knowledge for others to learn and Today, Africans draw on all these traditions in fighting use from it. While those who inherits by birth; not all illness and pursuing health. Several factors can explain knowledge captured are known. These contributes to the the increasing success of Tradi- tional medicine(TM): design of a knowledge based ontology model on ATM alternative for low income households, less restrictive which provides formal framework to maintain, and very often less expensive than the high cost of document, reuse and promote knowledge sharing for Western Medicine (WM). For this reason, many healthcare. Inorder to provide a formal framework for government officials are taken a closer and more serious knowledge shared conceptualism with terms and look at African Traditional Medicine (ATM) methods concepts through formal description of ATM, the and institution by WHO, to work with researchers by knowledge representation should be influenced by promoting the use of TM for healthcare in [8]. Unlike ontology [3]. According to [1], the first and only attempt the Traditional Chinese Medicine (TCM) which is well to build ontology for ATM was constructed in [6]. The formalized through research. Knowledge acquired in work demonstrates knowledge representation in most African countries are acquired through inheritance artificial intelligent however, this attempt lacks formal (birth) or trained for it. semantic constructiveness in ontology description.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The ontology described mainly contains terms and information [7]. Moreover, because of the meta-data of relations including: “documents” as a primary unit of the knowledge units, there is increase in the lexical description. However, do requires interpretation in conceptual schema still remains unclarified. Over the relation to connected knowledge and considered in years, efforts to control medical terminologies have ATM, the necessity if an in-depth semantics such as resulted in various standards medical vocabularies such model theory, of some concept and relations may unfold as: Unified Medical Language System (UMLS) the mode of description and intervention of the chain of however, the problem of addressing complexity of evidenced- based reasoning, but which can be captured, controlled medical vocabularies was encountered [9]. formalized and interpreted in the ontology. Interestingly, the parallel demand of uniformity of reference by Another category of developing ontology was practitioners of TM and the model theoretic introduced in [7] where he analyzed past frameworks interpretation requires for computational representation used in knowledge representation (i.e. static, dynamic) scheme have to be reconciled. This paper presents a then relates techniques to information science in the framework for describing and formalizing expressive domain of knowledge management research however, concepts which will reveal the nature of ATM better was unable to come to a standard in knowledge capturing compared to [1]. and formalization compared to the existing works. Ontologies may be constructed to enable information We provide a framework for the formal sharing such as: [2]: presents source of information as representation of the practitioners of ATM and some a thesaurus and oracle database. In this paper, concepts associated terms. In this paper, we present an approach are identified either by pattern extraction or associated to build an ATM, Knowledge-based Ontology Model rule or by concept clustering. The extracted words are (KBOM) with the capability of logic reasoning based built as machine readable dictionary and are converted on properties which gives knowledge holders the into a knowledge base but his works was unable to privilege to express their knowledge on ATM and also represent domain knowledge (entities and relationship) scientist (Information System Per- spective) to embed as a conceptualization explicitly. various scientific interpretations and proofs in relation with each classes described. Our new approach 2.2. Related Works in Ontology Modeling Tools constructs a “Knowledge- based ontology model” in Ontology modeling tools mainly emphasizes on ATM by focusing on different instances of its use. choosing a representation language that will support the Finally, the rest of this paper is organized as follows: ontology. Some of these ontology modeling tools have provide some brief related works in section 2, we been reviewed such as: existing work on [4] that motivate the need for its ontology in section 3. We identifies the instances and relationships of the web present Description Logics Inference (DLI) for reasoning elements and coded using XML. The XML schema is to formalize the ontology in section 4. Section 5 is visualized as a DOM tree. Alternatively, some system dedicated to the presentation of a formal framework to may not use code to indicate hierarchical location e.g. build a KBOM for ATM. In section 6, we describe the SNOMED.RT. Xu and Luo (2007) proposes a Knowledge ontology browser interface and section 7, methodology using Protégé 2000 with RDFS as the conclude how this work could be extended and reused. representation language to construct ontology for a medical knowledge base. The structure of the medical 2. RELATED WORKS ontology will be a combination of western medicine and In this section, we shall present a brief introduction to traditional medicine. Protégé is a knowledge based the existing works related to this works. development framework that offers classes, slots, facets and instances as the building blocks for 2.1 Related Works in Knowledge Building representing knowledge. Protégé [Gennari et al. Several ontologies in various domains have been (2002)] helps knowledge engineers and domain experts developed using one language or the other and for to perform knowledge management tasks. It includes varying purposes. This depends on extensibility and support for class and class hierarchy with multiple simplicity of the knowledge representation language inheritances, slots having cordially restrictions, default and the targeted application domain such as: values, inverse slots, meta-class and meta-class Developing ontologies that cover domain and hierarchy. application can be used not only to support system integration but also system development by reusing However, the work here was unable to explicitly the ontology.[5]: proposed that ontologies should deal formalize knowledge in the domain world The work with general assumptions concerning the explanatory proposed our paper is better off in that, we integrate invariant of a domain using: [identification and DLs as a formal language that incorporate conceptual measurement of the object along dimensions of modeling construct into a coherent logical framework. possible variations which will represent a natural It produces classes, attributes with domain and range evolution in the field of modeling] but limitation definitions between elements of ontology and its emphasizes on links between ontologies and instances.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

With Prot´eg´e we can create a semi-formal ontology, (all are stomach pains), examinations will first be for documentation and reusability. Coupled with a DL considered during questioning from the patient. rule module which will execute the rules generated by We know inu didun, majorly leads to typhoid (iba the rule module (inference engine) through the execution jedo jedo) and ulcer (ogbe inu). However, further interface to infer diagnosis and recommendations. questioning still will be carried out like: When did it start? . . . What have you applied? ...If yes, tell us your experience? If otherwise, cause is known, therapy 3. THE NATURE AND ATTRIBUTES OF ATMs for the cure can be in form of ipara(cream), agbo(liquid)and agunmu(powder). ATM is a body of knowledge that has been developed and accumulated by African thousands of years ago, concerned with examination, diagnosis, therapy, treatment, prevention and rehabilitation of the general 4. DESCRIPTION LOGIC INFERENCE wellbeing of humans [?] One of the important causal factors considered in ATM is the type of relationship The Description Logic based Inference (DLI) are that exist between the particular individual and other logic based formalism that provides automated human beings. In African philosophy, the human being support for the structured representation of terms, should not meet a reckless ‘sovereign ruler’ over nature. concepts and roles (elements of ontology) in a Traditional medicine and traditional healers from part of given domain (ref ). LBIs are an important area of a broader field of study classified by medical knowledge representation (KR) in artificial intelligence anthropologists as ethno medicine (Marlise, 2003), (Description Logics). The knowledge described in DL Traditional healers are classified into two categories: comprises of: those that serve the role of diagnostians (diviners) and those who are healers (herbalists).  A components that describes terminologies i.e. Vocabulary in ATM such as: terms Diviners provide a diagnosis usually through spiritual (representing set of objects) and roles means, while the herbalists then choose and apply  A component that contains assertions about relevant herbal medicines. The [10] estimated that up to hidden objects with re- spects to this 80 percent of the population in Africa makes use of vocabulary. traditional medicines. Traditional healers provide client- center, personalized healthcare that is culturally With atomic terms and roles, this DL allows complex appropriate, holistic and tailored to meet the needs and description of terms (class). We can summarize the expectations of the patient. Traditional healers are syntax and the semantics of some description culturally close to clients, which facilitates languages. communication about diseases and related social issues. 4.1 Proposed Model 3.1 Description of Ailments in Yoruba Our Model employs Description Logics which The Yoruba concept of the preventative and curative is provide adequate inference mechanism for part of the day-t-day existence of the people which is hierarchical concepts descriptions. The Description inferred by their health beliefs. According to Ayodele Logic has two phases: (2002), he observed that certain minor illness like iko eyin (cough accompanying growth of teeth), inu dodo • Vocabulary description. (spasm), and paanu (baby skin rash after birth) are normal conditions in child development. Illness is an • Hidden Assertion. aspect of aisan (not well) thus, illness is an abnormal phenomenon which requires corrective actions. For example, if we say: Inu rerun, ini kikun, and inu didun

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Description Logic make use of syntax Symbols:

∀(for all), ∃(quantifier), ¬ (not), ⊆ (subset), ∩ (intersection), ∪ (union), r (Restriction),j (interpreted), 3 (denotes), ≡ (multiple subset), a(implies) etc

Semantics: ∀∃R.∩-The set of all objects related through R with other objects in the domain R → rq

The inverse of role represented by Z

FOPterm ⊆ ∃ FOP ∩…...... (1)

If P and Q are terms:

P∩Q ⊆ (P ∪ Q) ∩ ∃R.P...... (2)

Description of P and Q will be related. j j ∗ Pi ⊆ P Oj = {Pi ⊆ P ; i = 1, ..., n, n ∈ N i i Therefore, if a term P Q can be interpreted as W i

e Pi ⊆ Nk = 1, ..., w(∃iA.p ) ...... (3)

Hidden ontology description, Aspect can be expressed as:

j P ⊆ P (N I ) ∩ (n = 1, ..., w(∃iA.P )) …………...... ………(4) i k i ik

5.ATM FORMALIZATION

Examples

1. “Collect fresh herbs, dry the herbs and soaked into cold water, take a cup of portion once a day”.

2. “herbs collected must be preserved in driedForm and soaked in cold water. Once the chemical has been extracted, patient must take a 5ml cup of solution twice a day for two weeks”. soaked ⊆ portion ∩ ∀M adeF rom.driedH erb (5)

Lexical description of the therapy has the following:

• Species/Part-used • Time-of-Collection • Preparation • Dosage • frequency • Restriction

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Extend term ”soaked” to ”form” to ”dosage” and ”restriction” soaked ⊆ N I ∪ ∃iA.Xd osage ∪ ∃iA.Xr estriction …………………….(6)

From equation 6 soaked ⊆ (portion ∩ ∀madeF rom.driedH erb) ……………….(7) ∩(N I ∪ ∃iA.dosage ∪ ∃iA.restriction) ……………………….(8)

Finally soaked ⊆ (portion∩∀madeF rom.driedH erb∩N I )∪(portion∩∀madeF rom.driedH erb∩∃iA.X –d ……………..(9)

Soaked denotes portion madeFrom driedHerbs which defines a certain dosage and constraint

The Hidden Assertion can be represented as follows: To construct

1. DriedH erb 3 P reservationF orm − spec 2. Soaked 3 status − spec 3. acup 3 quantity − spec 4. once − adayf or2weeks 3 f requency − spec

The description process will be soaked ⊆ status−spec∩(N I ∪∃iA.extract obtained)once aday ⊆ quantity spec(N I ∪∃iA.ef f ecti ……………..(10)

5.1 Application Area: edge, Relationship, Constraint and Attributes represented with the ontology ATM. The rule of inference coupled The ultimate aim of the ontology construction in this with DL rule draws conclusion based on the assertions work is to redefine ATM terminologies and to bring to interpreted and analyzed and returns a conclusion standard, ATM practices while initiating a knowledge while the inference engine been an expert system uses based application development. For example: the the rule of logics to infer decisions from the knowledge web-based ATM/HS will make it possible for software base. engineers to gain full access to information of patients and practitioners of ATM, with their beliefs and in- The justification trace module will be generated to digenous practices in order to into patient specific explain the proposed recommendations and the different recommendation and to further explain these interpretation while the Facet which runs on the recommendations. This also will help to ease down psy- description logic inference engine using the rule of chological consideration and trauma experienced by inference to infer diagnosis and recommendations. practitioners of ATM. At this level, the already existing Using the rule of inference, hidden aspects based on ATM/CDD fails because they do not consider beliefs and terms are described during diagnosis and therapeutic other meta-physics aspect. To simulate an interactive description in our proposed knowledge based ontology web based ATM on this ontology; we present an model. For example, a recommendation may include the alternative cure for medical diagnosis and treatment form “soaked” and briefly explain the condition behind it. recommendation. Finally, the explanation will satisfy the reason of the condition as a negative impact/effect if not soaked into The focus of our e-ATM/HS is the medical diagnosis water with respect to the type of disease and extent to and treatment recommendation including representing which it will be cured. knowledge coupled to a logic-based inference engine to reason and make decision over the ontology. We are 6. CONCLUSION using prot´eg´e-OWL coupled with rule of inference to infer logic-based recommendation over our ontology With the development of knowledge based ontology (web-based ontology language-Description logics). The model for ATM integrated with assertions (hidden rule module and execution module interface will aspects) and interpretations of the terms has surfaced that provide ATM experts an interface to specify decisions some instances/aspects considered metaphysics initially and to infer recommendations based of the decision sometimes hides both scientific and critical making. The rule made will provide ATM engineers an interpretations which can formally be represented in interface to specify decision logics guided by the knowl- ontology domain of ATM. For this, knowledge of the

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

concern domain has to be represented. In Africa, [4] Chen-Heui chou and Fatemeh Z. ”ontology for where knowledge and practices are been transferred developing websites for natural disaster through inheritance and master to disciple, which have management methodology and not been effective (un-formalized), formalizations are implementation”. 2008. of great importance therefore, inorder to document, reuse and promote knowledge sharing with ontology [5] Fonseca Frederico. “the double role of ontology in building tool: Protégé. In this paper, we have information science research”. In a Journal of provided a formal framework that enables domain the American society for Information Science experts to build domain knowledge in ATM using DL and Technology, USA., 58(6), 2009. Language to represent formally, knowledge, hidden aspects in ATM. [6] Atemezing G. and Pavon J. An ontology for african traditional medicine. In In International Symposium on Distributed Computing, With the improvements on the existing works [2] by Artificial In- telligence and Soft-Computing, incorporating hidden assertions associated to various Salmanca, volume 50, pages 329–337. 2008. terms used in ATM, knowledge sharing and interoperability of practices between the practitioners of [7] Igor Jurisica, Mylopolous J., and Yu E. Using ATM and the users (patients) will be achieved. Once the ontologies for knowledge management - an developed ontology have been formalized and built information system perspective. In the using Protégé-OWL, it can be used to build a re- Journal of Computer Science, University of defined ontology for ATM inorder to promote, Toronto, Ontario, Canada, pages 10–12, documenation, reusability and knowledge sharing among 1999. practitioners of ATM and thereafter, we parse the knowledge formalized in XML language into the [8] Adeshina S. K. Traditional medical care in knowledge base (database) while queries and responds nigeria. can be made through the portal/search interface thus, http://www.onlinenigeria.com/traditional this will be used to provide more accurate search in medical care nigeria.com/,2011. ATM domain, provides personalized search by consulting with the practitioners of the concerned [9] Grummer M. ”methodology for the design and discipline (fetish, healers, ethno botanists, herbal evolution of ontology”. In Workshop on Basic specialists) such as: eForum in ATM system. The Ontology Issues on Knowledge Sharing, framework also provides an improved architecture for Interna- tional Workshop on Open Enterprise our proposed knowledge based ontology model for ATM Solution: System, Experience and system while reviewing the existing works. Organization, Rome., 1995.

REFERENCES [10] World Health Organization. “traditional medicine strategy”. Fact sheet, World Health [1] Armel Ayimdji, Souleymane Koussoube, Laure P. Organization, Geneva, Switzerland, 2002. Fosto, and Balira O. Konfe. Towards a ”deep” ontology for african traditional medicine. In Journal of Intelligent Information Management, 3:244–251, 2011.

[2] De Quen Bing Ru, Yang and Zheng JY. ”automatic ontology construction approaches and its applications on military intelligence”. In In Pacific Conference on Information Processing, Asia., 2009.

[3] Diego C. and Luciano S. Knowledge representation and ontologies. In Journal of Computer Science, FREE University of BOZEN., 1, 2011.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Design and Implementation of an Enhanced Power Billing System for Electricity Consumers in Nigeria

Adegboye Adegboyega Department of Computer Science Achievers University, Owo, Nigeria [email protected]

Ayeni .A. Gabriel Department of Computer Science Allover Central Polytechnic Ota, Nigeria [email protected]

Alawode .J. Ademola Department of Computer Science. The Federal Polytechnic Ilaro, Nigeria [email protected]

Azeta .I. Victor Department. of Mgt, Labour and Productivity National Productivity Center Kaduna, Nigeri

ABSTRACT In Nigeria, electricity consumers are often faced with the problems of inaccurate, irrational and delay in monthly billing due to the drawback in reading pattern and human errors. Thus, it is essential to have an efficient and effective system for such purposes via electronic platform with consideration to proximity. This paper presents the design and functional significance of a web-based application with online capability called Power Billing System (PBS). PBS is a solution system developed with Microsoft Visual Web Development IDE; being an Object Oriented Design tool from Microsoft Visual Studio.net collection and Microsoft Access with SQL query for back-end database. It measures accurately the electric power consumed by residential or commercial buildings which is more economical compared to the electromechanical devices. Individual consumer and the utility companies can directly monitor and control electric power supply billing without engaging the services of meter readers. It displays the sale rate of electrical power per unit and the consumed power per minute. It provides environment to maintain the consumer details right from connection and performance information to the management. It is an Intranet and Internet based software solution that ensures timely availability of status parameters.

Keywords: Power Billing, Electronic System, Electricity, Meter Reading, Consumers, Database

African Journal of Computing & ICT Reference Format Adegboye Adegboyega, Ayeni .A. Gabriel, Alawode .J. Ademola & Azeta .I. Victor (2013). Design and Implementation of an Enhanced Power Billing System for Electricity Consumers in Nigeria. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 49-58

1.0 INTRODUCTION It provides an environment to maintain the consumer details starting from getting new connection, receiving Power Billing System is an Executive Information bill, payments etc; access to performance information by System (EIS) that determines the consumed power per the management (Seshanna et al, 2006). It functions on unit time and performs its computation based on the sale an Intranet network and Internet domain and ensure rate of power per unit time and other parameters. The timely availability of status parameters. The ability to importance of Power Billing System (PBS) cannot be view the reports online ensures access to the report from over emphasized because its calculation reflects the exact PC terminal or devices VLAN and WAN network with power consumption for the prospective consumers, and in internet connection. Customers can lodge complaint or monitoring the billing details of the electricity consumers deal with new connections just by logging into the (Advalorem, 2009). system.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

In the traditional system, files were used to maintain the The recognition of electromagnetism, the unity of electric database which was done manually. This existing system and magnetic phenomena; electricity and magnetism consumes a lot of time. This time consuming evaluation were eventually linked. The Power Holding Company coupled by the huge maintenance problem and also leads of Nigeria (PHCN), formerly the National Electric Power to erroneous results in most cases. The various operations Authority (NEPA) is an organization governing the use performed on these files by the personnel of Power of electricity in Nigeria. Despite the problems faced by Holding Company of Nigeria (PHCN) like sorting, NEPA, the authority has played an effective role in the adding, modifying and deletion of the records are very nation's socio economic development thereby steering tedious. Moreover, these manually maintained files have Nigeria into a greater industrial society. The success story the possibility of getting worn out. Thus, less durability, is a result of careful planning and hard work. The reliability, privacy, prioritization and efficiency is statutory function of the Authority is to develop and achieved. maintain an efficient co-ordinate and economical system of electricity supply throughout the Federation. 2.0 PROBLEM ANALYSIS The decree further states that the monopoly of all Electricity is the science, engineering, technology and commercial electric supply shall be enjoyed by NEPA to physical phenomena associated with the presence and the exclusion of all other organizations. This however, flow of electric charges. Electricity gives a wide variety does not prevent private individuals who wish to buy and of well-known electrical effects, such as lighting, static run thermal plants for domestic use from doing so. electricity, electromagnetic induction and the flow of NEPA, from 1989, has since gained another status-that of electrical current in an electrical wire (IEEE, 2008). In quasi-commercialization. By this, NEPA has been addition, electricity permits the creation and reception of granted partial autonomy and by implication, it is to feed electromagnetic radiation such as radio waves. In itself. The total generating capacity of the six major electricity, charges produce electromagnetic fields which power stations is 3,450 megawatts. In spite of act on other charges (Franklin, 1869). Priestley (1967) considerable achievements of recent times with regards to Electricity remained little more than an intellectual its generating capability, additional power plants would curiosity for almost a millennium until a careful study of need to be committed to cover expected future loads. electricity and magnetism, distinguishing the lodestone effect from static electricity produced by rubbing amber At present, plans are already nearing completion for the (Bryon, 2002). Alessandro Volta's battery, or voltaic pile, extension and reinforcement of the existing transmission of 1800, made from alternating layers of zinc and copper, system to ensure adequate and reliable power supply to provided scientists with a more reliable source of all parts of the country. The existing system is a billing electrical energy than the electrostatic machines machine that constitutes five divisions but, too previously used (Abubakar, 2009). overburdened, less flexible, slow pace of processing and not so user’s friendly. Fig. 1 gives the overall block diagram of the computer-based power billing machine with the highlighted shortcomings. This is an electromechanical meter system used to measure accurately the

electric power consumed by a company or an individual.

Power Supply Unit

Power Analog to Digital Computer Switch Converter Converter (ADC) Parallel Port Control

Fig. 1 Block diagram of power billing machine

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2.1 Consumers Classification Consumers are classified based on their mode of consumption and by tariff classification. Tariff measures; defined as electric tariff at which the energy is selling to the consumers (Ghoshal, 1997). Usually electricity tariff are fixed by Government. Tariff at the moment are categorized into residential, commercial, industrial, street light and special tariff. The special tariff is agro-allied enterprises, Government and teaching hospitals, water boards, secondary and tertiary Institutions. The tariff for each category is fixed by voltage class (Abubakar, 2009). Tariff is calculated by kilowatt hour. For industrial and commercial or other consumers, receiving transformer with a capacity of 100 KVA or more, who have electrical equipment installed, receiving capacity of 100 KW or more; their tariff comprises two components, these are Kilowatt hour tariff (calculated on the basis of actual use) and basic electricity tariff (based on the consumption capacity).

Table 1: Residential Class Class Demand Level Demand Minimum Fixed Meter Energy

Charge/Kva Charge/Month Charge Maintenance Charge/Kwh

Charge

R1 < 5kVA 0.00 31.00 31.00 154.00 1.30

>= 5 < 0.00 46.00 46.00 154.00 4.40

R2 15KVA

R3 >= 15 < 0.00 185.00 185.00 772.00 6.60

45kVA

R4(MD) > 45<500kVA 0.00 7,716.00 185.00 2,469.00 9.40

R5(MD) >=500<2MVA 0.00 48,228.00 0.00 3,395.00 9.40

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Table 2: Industrial Class Class Demand Level Demand Minimum Fixed Meter Energy Charge/Kva Charge/Month Charge Maintenance Charge/Kwh Charge D1 >5<15kVA 00.00 136.00 136.00 151.00 7.90 D2 >15<45kVA 00.00 181.00 181.00 755.00 10.30 D3 >45<500kVA 278.88 7,550.00 362.00 2,416.00 10.30 D4 >500<2MVA 303.13 47,188.00 0.00 3,322.00 10.30 D5 >2MVA 327.38 2,265,011.00 0.00 3,322.00 10.30

Table 3: Commercial Class Class Demand Level Demand Minimum Fixed Meter Mainteanance Energy Charge/Kva Charge/Month Charge Charge Charge/Kwh

C1 >5<15kVA 00.00 138.00 138.00 153.00 7.40 C2 >15<45kVA 00.00 184.00 184.00 767.00 9.70 C3(MD) >45<500kVA 262.53 7,673.00 368.00 2,456.00 9.70 C4(MD) >500<2MVA 32813 47,959.00 0.00 3,376.00 9.70

Table 4: Street Lighting Class Class Demand Level Demand Minimum Fixed Meter Maintenance Energy Charge/Kva Charge/Month Charge Charge Charge/K wh S1 1-PH, 3-PH 0.00 312.00 0.00 651.00 5.90

3.0 METHODOLOGY The major specification in this design allows the system Formal model of the proposed system is presented in to capture data related to consumer’s profile in order to flowchart and context diagrams. All these models will assign an identification code with which transaction give the conceptual view and to provide the graphical relating power billing, meter request, complaint. It analysis of users’ requirements. As a major modeling constitutes various modules among which administrator tool, entity relationship diagrams helped in organizing the and consumer module are integral. Consumer is granted functional elements of the system into entities and also access only through the username and password created define the relationships between the entities. This process from first visit to online system or when registering at the enabled the analyst to understand database structure so web portal, in orders to utilize the features available to that data can be stored and retrieved in a most efficient consumer from remote terminal. The administrator manner. Flowchart showed the flow of data from external module is handled by an authorized PHCN employee, in entities into the system. It also showed how data moved order to grant request relating to customers’ service, from one process to another as well as its logical storage. validate every transaction online or to confirm payment Figure 2 shows the operational modalities that guides via electronic system and to review consumers’ profile, input-output process via users’ interface while Figure 3 revert an action, track connection and billing status. The shows the major activities of the consumers and functional requirements of this application were analyzed integrated into web portal and as online system for from the data contained in the existing system; inputs like electricity supply and bill distribution, and as well as the business information, service stations and the data automated interactivity of the consumers’ module in contained in the outputs like the bills, ledger and receipt. validating users’ input.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Start

Read Uname,

Pwd

False Pwd and Uname True Is Fig. 2 Login Flowchart Status = Admin ? correct ? Start True

False

Admin Consumer home Page Error message

Consumer Page

Consumer View Profile HomepageEdit Details Post My Billings Log out home menu menu Query menu menu

Stop

View bill View Edit Query and payment Consumer Consumer

Homepage

Stop

Fig. 3 Consumers’ Activity Flowchart

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4.0 IMPLEMENTATION AND RESULTS The implementation was done using Microsoft Visual Web Developer and Access Database with SQL support for back- end application. The System captures Information related to actual demand, energy usage, payments, exceptions etc. from various levels of organization with the aim of capturing it from as close to the source as possible. The application basically starts by displaying the homepage. Thereafter, username and password are requested for the specified status (i.e. consumer or administrator). Validating the username and password in order to proceed or not. During the integration testing, the following outputs were obtained according to design modules; in order to describe and analyze the functional scope and performance evaluation of Power Billing System (PBS).

1. Meter: Enables the administrator to add and view consumer meter by providing the circle name, division name, meter company name and the meter I.D, If the meter I.D matches with another I.D in the database, it will prompt an error indicating that the I.D. cannot be used.

Fig. 4: Meter Registration Screen

Fig. 5: Consumer Home Screen

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2. Billings: Enables the administrator to send bills to consumer by selecting the connection type and consumer meter number. Note, consumer cannot be billed more than once in a month.

Fig. 6: Select Connection Screen

3. My Billing: This session enables user to make bill payment by providing a card number. If this card number is invalid then the system will prompt an error via merchant and EPS support.

Fig. 7: Payment Screen

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4. Post Query: Allows consumer to post complaint to the administrator by clicking on the Post Query menu and to type their query or comment in the textbox provided on the page.

Fig. 8: Post Query Screen

5. Admin Login Session: Enables the administrator to have access to the system via username and password. If the password or username entered is wrong, then the system will automatically display an error page indicating that the username or password is incorrect.

Fig. 9: Admin Login Screen

6. View Query: Enables the administrator to view consumer complaint. This is achieved by clicking on the View Query menu.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Fig. 10: View Query Screen

5.0 CONCLUSION 6. REFERENCES

Usability testing was part of post implementation Advalorem, U.B (2009). Utility Billing Software: review and performance evaluation for Power Billing Energy Billing System. [Online] System (EPS), in order to ensure that the intended users Retrieved from <” http:// of the newly developed system can carry out the www.avrub.com/contact.htm16”> on 24 /12/ 2012. intended tasks effectively using real data so as to ascertain the acceptance of the system and operational Abubakar, S.M (2009). Design and Construction of a efficiency. It caters for consumers’ bills and also Computer Based Power Billing System. enables the administrator to generate monthly reports. (Published Dissertation), Federal University of It is possible for an administrator to know the Technology, Minna, Nigeria. AU J.T. 13(1): 39-46 consumers that have made payment in respect of their bills for the current month, thereby improving the Byron, G.S (2002). Shaum’s Outline of Theory and billing accuracy, reduce time consumption and Problems of Programming with Visual workload on PHCN employees or designated staff, Basic: Essentials of Visual Studio.net. New York, NY; increase the velocity of electricity distribution, Mc Graw Hill Incorporation. connection, tariff scheduling, eliminates variations in bills and replenish based on market demand. The Ghoshal, K. (1997). Distribution Automation: SCADA conceptual framework allows necessary adjustment and Integration is the key. IEEE Journal of Computer enhancement maintenance to integrate future demands Applications in Power and Control Systems. Vol. 2, according to technological or environmental changes Issue 1, Pp. 31-38. with time. It manages the consumers’ data and validates their inputs with immediate notification to Hall, D.V (1992). Micro Processors and Interfacing - users at remote locations; centralized in PHCN offices Programming and Hardware (Second Edition) - across the nation. Collaborative Design, Singapore; McGraw Hill International. IEEE (2008). Principles and Practice in Electricity Metering. (Quarterly Bulletin of the Institute of Electrical / Electronic Engineers) Retrieved from < www.ieeeexplore.com> on 25/01/2013.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Misner, S. & Mistry, R. (2012). Introducing Microsoft SQL Server 2012 (Microsoft Team And Consortium Release) . Washington, D.C; Microsoft Press International. ALAWODE, J. Ademola is a lecturer in the Seshanna, P.; Nashtara, I.; and Sajeed, A.H (2006). Department of Computer Automated Industrial Load Measurement System. AU Science, Federal J.T. 10(1): 23-8. Polytechnic, Ilaro, Nigeria. He holds B.Tech in Computer Science from Authors’ Biography Ladoke Akintola University of Technology, Ogbomoso; and M.Sc in Computer Science (In View). He also obtained professional certification in Enterprise ADEGBOYE, Adegboyega is Application Development with Advanced Java a lecturer in the Department of Programming. He has contributed articles in Local and Computer Science, Achievers International Journals with publication of Educational University, Owo, Nigeria. He texts. His current research interest includes Software holds B.Sc, MBA, M.Sc and Development and Engineering Metrics, Data Mining Ph.D (In View) in Computer Algorithm and Database Security. He is a member of Science from Ogun State Nigeria Computer Society. University now Olabisi Onabanjo University, Ago Iwoye; University of Ado Ekiti now Ekiti State University, Nigeria; Federal University of Agriculture, Abeokuta respectively. In AZETA, I. Victor is a staff of complement to OND in Electronic/Electrical Department of Management, Engineering from Yaba College of Technology, and Labour and Productivity HND in Electrical Engineering (Power Option) from (MLP), National Productivity The Polytechnic Ibadan. His current research interests Center, Kaduna State office, spans through Software Metrics, Data Mining and E- Kaduna. He holds a B.Sc Governance Framework. He is a member of Nigeria Political Science and M.Sc Computer Society (NCS), Computer Professional Public Administration from Edo (Registration Council) of Nigeria (CPN), Nigeria State University and University Society of Engineer (NSE), Council Regulating of Calabar respectively. The current research interests Engineering in Nigeria (COREN), and Association of include: Productivity, management, Administration and Maintenance and Design Engineers (MADE). Finance, e-Government and Leadership style.

AYENI, A. Gabriel is a lecturer in the Department of Computer Science, Allover Central Polytechnic, Ota, Nigeria. He

holds B.Sc in Computer Science from University of Ado Ekiti, Nigeria now Ekiti State University, and M.Sc in Computer Science (In View). He also obtained professional qualifications in Computing and Information Technology such as MCP, A+, N+, CCIA, CCHA, OCA-SQL among others. He has published Journal articles and educational texts. His current research interests are in the following areas: Software Engineering, Network Security, Web Metrics, Cloud Computing, Algorithm Design and Data Mining. He is a member of the Nigeria Computer Society (NCS), Oracle Technology Network (OTN), Microsoft Virtual Academy (MVA) and Association of Computing Machinery (ACM).

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Better Quality of Service Management With Fuzzy Logic In Mobile Adhoc Network

Onifade O.F.W. Department of Computer Science University of Ibadan Ibadan, Nigeria

Ojesanmi O.A. Department of Computer Science Federal University of Agriculture Abeokuta, Nigeria.

Oyebisi T.O. African Institute of Science Policy and Innovation Obafemi Awolowo University Ile-Ife, Nigeria. [email protected]

ABSTRACT Quality of service (QoS)is a great concept in mobile Adhoc network (MANETs). It is of a great importance that we are very conscious of how packet is routed to maximize efficiency and minimize delay. In this paper, an efficient algorithm for transmitting packet for better quality of service in adhoc mobile network was proposed. Fuzzy Self Organizing Map (FSOM) provide very efficient algorithmic tools for transmitting packet in an efficient manner by taking the most efficient route and also the bandwidth, latency and range are considered to determine how good is the data delivered. The results shown that fuzzy logic can guarantee QoS of every packets inthe network.

Keywords – QoS, Adhoc network, Packet, Fuzzy logic.

African Journal of Computing & ICT Reference Format Onifade O.F.W.,Ojesanmi O.A. & Oyebisi T.O. (2013). Better Quality of Service Management With Fuzzy Logic In Mobile Adhoc Network. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 59-68

1. INTRODUCTION

The earliest mobile ad-hoc networks were called “packet In the past few years, we have seen a rapid expansion in radio" networks, and were sponsored by DARPA in the the field of mobile computing due to the proliferation of early 1970s [3]. Then, the advantages such as flexibility, inexpensive, widely available wireless devices. However, mobility, resilience and independence of fixed current devices, applications and protocols are solely infrastructure, elicited immediate interest among military, focused on cellular or wireless local area networks police and rescue agencies in the use of such networks (WLANs), not taking into account the great potential under disorganized or hostile environments. For a long offered by mobile ad hoc networking. A mobile ad hoc time, ad hoc network research stayed in the realm of the network is an autonomous collection of mobile devices military, and only in the middle of 1990, with the advent (laptops, smart phones, sensors, etc.) that communicate of commercial radio technologies, did the wireless with each other over wireless links and cooperate in a research community became aware of the great potential distributed manner in order to provide the necessary and advantages of mobile adhoc networks outside the network functionality in the absence of a fixed military domain, witnessed by the creation of the Mobile infrastructure, with no central server, self organizing Adhoc Networking working group within the IETF [18]. networks [9].

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The routers are free to move randomly and organize The works in [4][5][6][12][15][16][18][21][23][25] have themselves arbitrarily, and thus, the network's wireless studied the QoS routing issue. Most of the existing topology may change rapidly and unpredictable. Manets distributed algorithms (e.g., [23]) require the maintaining has the chance of potential ease of deployment, decreased of a global network state at every node, which may cause dependency on infrastructure [11][14]. The majority of the scalability problem. On the other hand, the source applications of MANets are in areas where rapid routing schemes such as [5] suffer from problems of deployment and dynamic reconfiguration are necessary scalability and frequent updates of the state of the and wired network is not available. They are applicable network. Hence the need to have a scheme that in universities, rescue sites, classrooms and conventions guarantees good QoS for the packets. where freedom of location, mobility may be of paramount importance and where participants share The paper is organized as follows. Next section discusses information dynamically using their mobile devices [9]. our architecture for quality of service management adhoc network using fuzzy logic. Section III presents the Applications of mobile ad hoc networks have increased simulation study and performance evaluation, followed requirements in order to ensure high quality of service for by conclusion in section IV. the provided services and all these identified areas of application need quality of service for their connection 2. SYSTEM ARCHITECTURE lifetime [8][26]. Manets are facing constraints such as: 2.1 Overview of the architecture limited bandwidth, power constraints, mobility, dynamic A system is a collection of various component which topology, high link error rate, Ease of snooping on interact with one another in a manner that satisfy wireless transmissions (security hazard), multi-hop objective and goals according to a set of functional and routing and network scalability makes quality of service performance specifications. There are two stages of in Manets a challenging task. In Mobile Adhoc Networks design which is the system design and program design. (MANETs), determining the physical location of nodes The system design is the stage of system development (localization) is very important for many network which determine what the proposed system will do and services and protocols by [10][22]. how it would be done. While the program design deals with the mechanisms that best implements the solution. The QoS satisfaction problem in wireless ad hoc network System design can be described as either a product or as a has been studied by many researchers. Recently, there process. As a product, it is viewed as a product resulting has been research in the area of supporting QoS in in the transformation of the problem into solution. On the MANETs. [1] did a study on qos provision for IP-based other hand, system design is two step process: the data radio access networks.[2] described a model towards model part and the process model part. achieving QoS guarantees in Mobile Ad hoc Networks. Process modeling: it shows how data flows among [24], [13] and [19] proposed methods of supporting different processes and how they transform. quality of service in mobile ad hoc networks. [17] describes a semi-stateless approach based on a fuzzy 2.2 Network Learning Algorithm logic system for wireless mobile ad hoc networks. The In neural network structure, each output neuron directly works that exist tend to be based on distributed corresponds to a class of trajectories. The number of scheduling algorithms that address QoS routing issue, output neurons used to describe the activity patterns is QoS-based medium access controllers, rescheduling essentially arbitrary. The more neurons used the greater when the network topology changes, and fairness issues. the accuracy of the model. The number of output neurons needed for good accuracy depends on the complexity of a scene. The more complex the scene is, the more output neurons are required. The number of output neurons is manually selected. The weights (W) connect the input vector components and the output neurons.

The weight vectors are of the same dimensions as the sample vectors. The weight components are initialized randomly and adjusted gradually using a self-organizing learning algorithm, and ultimately a mapping, from input to output, that keeps the distribution features of trajectories formed [15].

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Let M denote the number of input samples, N the number If the stability condition is satisfied or the predefined of input vector components and K the number of output number of iterations is achieved, then the learning neurons. The learning algorithm consists of the following process terminates; otherwise go to Step 2 for another steps. loop of learning. From the above learning procedure, we can see that the fuzzy SOM eases the difficulty of Step 1 selecting network parameters. The weights are adjusted Randomize the initial values of the components of the only once in each learning loop and the features of all weight vectors. input samples are taken into consideration once the weights are adjusted, so the learning speed and estimation Step2 accuracy are both greatly improved. In fact, different Input all samples kinds of fuzzy membership functions can be used in the above learning algorithm.

l = 1,2; ... The generation of fuzzy membership function via SOFM ;M. has, so far been a two-step procedure. The first step

generates the proper clusters. Then, the fuzzy Step 3 membership function is generated according to the Calculate the Euclidean distances from each sample Xl to clusters in the first step. However, it is possible to all output neurons integrate the two-step procedure and generate the fuzzy

membership function directly during the learning phase. The main idea is to augment the input feature vector with the class labeling information. The variables associated are semantic with the objective to cluster and visualize the data distribution. The focus was on how SOFM could be used to handle fuzzy information. Therefore, the information being associated Step 4 are all fuzzy variables [7]. Compute the memberships of each sample to all neurons 3. THE PROPOSED SYSTEM

The fuzzy SOM introduces the concept of membership function in the theory of fuzzy sets to the learning process in the batch manner. The membership of each sample to each neuron is calculated, and then the weight vector of each neuron is adjusted according to all the memberships of all samples to the neuron. In the fuzzy

SOM, some network parameters related to the

neighborhood in the SOFM are replaced with the Step 5 membership function. So the burden of choosing network Adjust the weights of each neuron according to the parameters is eased. Integration of the SOFM and other computed memberships fuzzy set based algorithms can produce other variants of

fuzzy SOM.

3.1 Fuzzy Inference Structure Model The FIS Editor, figure 1, handles the high level issues for the system: How many input and output variables? What are their names? The Fuzzy Logic Toolbox doesn’t limit the number of inputs. However, the number of inputs may be limited by the available memory of your Step 6) Determine the stability condition of the machine. If the number of inputs is too large, or the network number of membership functions is too big, then it may also be difficult to analyze the FIS using the other GUI tools. The Rule Viewer and the Surface Viewer are used for looking at, as opposed to editing, the FIS. They are strictly read-only tools.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The Rule Viewer is a MATLAB-based display of the 3.2 Fuzzy Logic System Analysis fuzzy inference diagram shown at the end of the last Fuzzy inference is the process of formulating the section. Used as a diagnostic, it can show (for example) mapping from a given input to an output using fuzzy which rules are active, or how individual membership logic. The mapping then provides a basis from which function shapes are influencing the results. The Surface decisions can be made, or patterns discerned. A fuzzy Viewer is used to display the dependency of one of the inference system is created based on the known outputs on any one or two of the inputs—that is, it sensitivity algorithm parameters to bandwidth, latency, generates and plots an output surface map for the system. range and this utilizes the Mamdani Fuzzy logic System. The Membership Function Editor is used to define the The singleton fuzzifier, the product operation fuzzy shapes of all the membership functions associated with implication for fuzzy inference, and the center average each variable. defuzzifier will be used.

The parameter depending on their availability are fed into a fuzzifier in which are converted into fuzzy sets. A fuzzy set contains varying degree of membership in a set. The membership values retrieved for a particular variable into a membership function (see figure 2).

Fig 1: Fuzzy Inference System Model

The steps involved in the design are:

Rule

Fuzzification Evaluation Aggregation Defuzzification Output

Figure 2: Stages in Fuzzy inference system

3.2.1 Fuzzification This is the process of generating membership values for a Membership function is designed for each quality of fuzzy variable using membership functions. The first step service condition which is a curve that defines how each is to take the crisp input variables and determine the point in the input space is mapped to a membership value degree to which these inputs belong to each appropriate (or degree of membership). fuzzy set. This crisp input is always a numeric value limited to the universe of discourse. Once the crisp inputs Input Variables with their value range are obtained, they are fuzzified against appropriate Bandwidth [0,10] linguistic fuzzy sets. Latency [0,10] Range [0,10]

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Output Variables with their value range 3.2.3 Aggregation Rule Output Data delivery [0,30] This is the process of unification of the outputs of all rules. In other words, we take the membership functions 3.2.2 Rule Evaluation of all the rules consequent previously scaled and combine This is the second step where the fuzzified inputs are them into single fuzzy sets (output). Thus, input of the applied to the antecedents of the fuzzy rules. Since the aggregation process is the list of scaled consequent fuzzy rule has multiple antecedents, fuzzy operator (AND membership functions and the output is one fuzzy set for or OR) is used to obtain a single member that represents each output variable (Name=’data delivery’ the result of the antecedent evaluation. We apply the Type=’mamdani’, Numinputs=3, NumOutputs=1, AND fuzzy operation (intersection) to evaluate the NumRules=’ ‘) conjunction of the rule antecedents. Rules added to this system are derived by mapping the three inputs to one 3.2.4 Defuzzification output by using conjunction (AND). Examples of some This is the last step in the fuzzy inference process, which of the rules are: is the process of transforming a fuzzy output of a fuzzy inference system into a crisp output. Fuzziness helps to (1) If (bandwidth is low) and (latency is low) then evaluate the rules, but the final output this system has to (data delivery is poor) (1) be a crisp number. The input for the defuzzification (2) If (bandwidth is high) and (latency is low) then process is the aggregate output fuzzy set and the output is (data delivery is excellent) (1) a number. This step was done using centroid technique (3) If (bandwidth is high) and (latency is low) and because it is most commonly used method of (range is very close) then (data delivery is Defuzzification (DefuzzMethod=’centroid’). excellent) (1) (4) If (bandwidth is low) and (latency is high) and (range is far) then (data delivery is poor)

The fuzzy logic-based resource management modeling architecture is given in figure 3.

Bandwidth

AND Latency Centroid Data

Rule

(min) technique Delivery Range

Block Figure 3: Architecture for fuzzy logic based Qos modeling.

4. EVALUATION We can infer from figures 4 and 5 that the time of delivery varies based on the route taken to transmit packet. We can conclude from figure 6 that the time and bandwidth for figure 4 is better than that of figure 5.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 4: Interface for the simulation Figure 6: Surface viewer from the of the best route. simulation interface

FIS Editor: Figure 7 is the window through which a new FIS type with any particular model can be selected, variable can be added, and input or output variable names can be changed. In this case, the chosen model is Mamdani. The pop-up menus in front of (And method to Defuzzification) are used to adjust the fuzzy inference functions, such as the defuzzification method. Double- click on an input or output variable icon to open the membership function editor, or on the system diagram to open the Rule Editor.

Figure 5: Interface for the simulation of the alternative route.

Figure 7: Fuzzy Inference System Editor

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Membership function editor: Figure 8 is the window through which the input or the output of the membership function can be changed and membership function can be added or removed. The graph field displays all the membership functions of the current variable. Click on a line in the graph and it is possible to change any of its attributes, including name, type and numerical parameters. The pop-menu right in front of type lets you change the type of the current membership function. The status line describes the most recent operation.

Figure 9: Rule Editor

Rule Viewer: The rule viewer in figure 10 displays a roadmap of the whole fuzzy inference process. It shows a graphical representation of each of the variable through all the rules, a representation of the combination of the rules, and a representation of the output from the defuzzification. It also shows the crisp value of the system. Data are entered for analysis through the rule viewer at the input text field.

Each column of plots (yellow) shows how the input variable is used in the rules. The input values are shown at the top, and the column of the plots (blue) shows how Figure 8: Membership function specification the output variable is used in the rules. Sliding the red for Qos line changes your input values, and generate a new output response although the edit field allows you to set the Rule editor: Figure 9 is used to add, change or delete input explicitly. The last output plot, a blue triangle rules, as the name implies. The rules are entered having a red line in between, the red line provides a automatically using the GUI tools. It provides defuzzified value, while the plot shows how the output of opportunity to change the connections and weight applied each rule is combined to make an aggregate output and to the rules (the default is always 1). Connection, link then defuzzified. input statements in rules. The (Delete, Add, Change rule) Create or edit rules with the GUI buttons and choices from the input or output selection menus and the not negate input or output statements in rules.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The surface viewer can generate a three-dimensional output surface where any two of the inputs vary, but two of the inputs must be held constant since computer monitors cannot display a five-dimensional shape. In such a case the input would be a four-dimensional vector with NaNs holding the place of the varying inputs while numerical values would indicate those values that remain fixed, A NaN is the IEEE symbol not number.

Figure 10: Rule Viewer for Qos

Surface Viewer: Figure 11 is a three dimensional curve that represents the mapping of three input. This is a three- input one-output systems as they generate three- dimensional plots that MATLAB can adeptly manage. When we move beyond three dimensions overall, we start to encounter trouble displaying the results. Accordingly, the surface viewer is equipped with pop-up menus that let you select any two inputs and any one output for plotting. Just below the pop-up menus are two text input fields that let you determine how many x-axis and y-axis grid lines you want to include. This allows you to keep the Figure 11: Surface Viewer calculation time reasonable for complex problems.

Clicking the evaluate button initiates the calculation and CONCLUSION the plot comes up soon after the calculation is complete. Fuzzy Self Organizing Map (FSOM) has been developed To change the x-axis or y-axis grids or Y-grids, in this paper. Incorporation of fuzziness in the input and according to which text field you changed, to redraw the output of the proposed model was seen to result in better plot. The surface viewer has a special capability that is performance. It should be noted that input variables are very helpful in cases with two (or more) inputs and one only three properties i.e. low, normal, and high were used output: you can actually grab the axes and reposition and for the output the variables are poor, good and them to get a different three-dimensional view on the excellent. data. The Ref. input field is used in situations when there are more inputs required by the system than the surface is mapping. Suppose you have a five-input one-output system and would like to see the output surface.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFERENCES

[1] Alberto L., Jukka M., Andrej M., Hector V., [11] Jukka M., Alberto L., Andrej M., Hector V., Eleanor H., and Youssef K. (1999), “A study Eleanor H., and Youssef K. (2002), on qos provision for ip-based radio access “Evaluation of Mobility and Qos Interaction”, networks”, IST project IST-1999-10050 Elsevier Journal of Computer Networks, 38, BRAIN. pp. 137–163.

[2] Arora, H. and Shetu, H. (2003). October [12] Khoukhi L., Cherkaoui S. (2004), “Flexible “Towards achieving QoS guarantees in Mobile QoS Routing Protocol for Mobile Ad Hoc Ad hoc Networks”, M.Sc Thesis, Faculty of Networks”, In Proc. of the 11th IEEE Drexel University. International Conference on Telecommunication (ICT2004), Brazil. [3] Carlos D. and Agrawal D.P. (2006) “Adhoc & Sensor Networks: Theory and application’, 2nd [13] Kui W. and Janelle H. “Qos support in mobile Edition, World Scientific Publishing Company, ad hoc networks”, Department of Computing pp. 1-641. Science, University of Alberta,

[4] Chansu Y., (2007).” Routing in Mobile Adhoc [14] Kurt G. (2002).” Analysis of adaptation Network”, Cleveland State University. strategies for mobile qos-aware applications”, M. Sc Thesis, Drexel University. [5] Chen S. and Nahrstedt K. (1998), “On finding multi-constrained paths”, IEEE International [15] Liao W.-H., Tseng Y.-C., Sheu J.-P., and Wang Conference on Communication, pp. 874 -879. S.-L. (2001), “A Multi-Path QoS Routing Protocol in a Wireless Mobile AdHoc [6] Chen S. and Nahrstedt K. (1999), “Distributed Network”, In Proc. of IEEE ICN’01, Quality-of-Service in Ad Hoc Networks”, IEEE International Conference on Networking, Part Journal on Selected Areas in Communications, II, pp. 158–167. vol.17, no. 8. [16] Lin C. R. and Liu J.-S.. (1999), “QoS Routing [7] Chih-Chung Y. and Bose, N.K. (2004), in Ad Hoc Wireless Networks”, IEEE Journal “Generating fuzzy membership function with on Selected Areas in Communication, Vol. 17, self-organizing feature map” Department of No. 8, 1426–1438. Electrical Engineering, Spatial and Temporal Signal Processing Center, the Pennsylvania [17] Lyes K. and Soumaya C. (2008), State University, USA. “Experimenting with Fuzzy Logic for QoS Management in Mobile Ad Hoc Networks”, [8] Demetris Z. (2000). “A Glance at QoS in International Journal of Computer Science and Mobile Ad-Hoc Networks (MANETs)”, Network Security (IJCSNS), Vol.8, No.8, pp. Department of Computer Science, University 372-386. of California, http://www.cs.ucr.edu/~csyiazti/cs260.html [18] Munaretto A., Badis H., Al Agha K. and Pujolle G. (2002), “A Linkstate QoS Routing [9] Hoebeke, J., Moerman, I., Dhoedt, B., and Protocol for Ad Hoc Networks”, In the Demeester, P (2005), “Towards adaptive ad proceedings of IEEE MWCN02, Stockholm, hoc network routing”, International Journal of Sept.2002. Wireless and Mobile Computing: Special Issue on ‘Wireless AdHoc Networking’ to be [19] Patrick S., (2003). “Quality of Service for published. Mobile Ad Hoc Networks”, Diploma Thesis, Eidgenossiche Technische Hochschule, Swiss [10] Jeroen H., Ingrid M., Bart D. and Piet D., “An Federal Institute of Technology, Zurich. Overview of Mobile Ad Hoc Networks: Applications and Challenges”, Department of [20] Perkins, C.E. and Royer, E.M. (2000). Quality Information Technology (INTEC)”, Ghent of Service for Ad Hoc on Demand Distance University – IMEC, Belgium. Vector Evaluation of the insignia signaling system. For ad hoc wireless networks (AODV) Routing, IETF Internet Draft, draft-ietf-manet- aodvqos-00.txt.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

[21] Sivakumar R. et al. (1999), “CEDAR: a core extraction distributed ad hoc routing algorithm”, IEEE Journal on Selected Areas in Communications, vol. 17, no. 8, pp. 1454-1465.

[22] Tinh, P. D., Noguchi, T. and Kawai, M. (2009).” Distributed Range-Free Localization Algorithm Based on Self-Organizing Maps”, Graduate School of Science and Engineering, Ritsumeikan University, Japan.

[23] Wang Z. and Crowcroft J. (1996), “QoS routing for supporting resource reservation”, IEEE Journal on Selected Areas in Communications, vol. 14, no. 7. http://www.meshnetworks.com.

[24] Xiao H., Seah W. K.G., Lo A. and Chua K. C. (2000), “Flexible Quality Service Model for Ad-HOC Networks”, In IEEE Vehicular Technology Conference, pp. 445–449, Tokyo, Japan. htpp://www.ece-icr.nus.edu.sg/journal1/ fqmmhandbook02.pdf.

[25] Xue Q. and Ganz A. (2003), “Ad hoc QoS on- demand routing (AQOR) in mobile ad hoc networks”, Journal of Parallel and Distributed Computing, Elsevier Science, USA.

[26] Zeinalipour-Yazti D., (2001) “A Glance at Quality of Services in Mobile Ad-Hoc Networks,” Seminar in Mobile Ad hoc Networks, Department of Computer Science, University of California-Riverside, USA.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Evaluating Usability Factors In Different Authentication Methods Using Artificial Neural Network

Mapayi, T. Department of Computer Science University of KwaZulu-Natal Durban, South Africa [email protected]

Olaniyan, O. M. & Isamotu N. O. Department of Computer Science Bells University of Technology Ota, Nigeria [tayoolaniyan2001; nehehandsome } @yahoo.com

Raji Shakirat.O. Department of Information Systems Faculty of Info. & Com. Technology International Islamic University Malaysia. [email protected]

Olaifa Moses S. Dept of Computer Systems Eng. School of Info. & Com. Technology Tshwane University of Technology Pretoria, South Africa [email protected]

ABSTRACT The human factor is often described as the weakest part of a security system and users are often identified to be the weakest link in the security chain. Likewise, authentication is a cornerstone of most security systems today, and most users interact with these mechanisms on a daily basis. Usability of the authentication mechanisms has seldom been investigated as a result little has been said about suitable evaluation model that considers usability and security. Over the years it has proved extremely demanding to merge usability with security in the choice of authentication methods. This somewhat mutual exclusivity of the two terms has placed users of these authentication methods in perilous positions.This research work develops a model that can help to ascertain the usability of the different authentication methods using artificial neural network.

Keywords: Authentication, Usability, Human Factors, Security Chain, Artificial Neural Network..

African Journal of Computing & ICT Reference Format Mapayi, T, Olaniyan, O.M, Isamotu, N.O., Raji, S.O. & Olaifa, M.S. (2013). Evaluating Usability Factors In Different Authentication Methods Using Artificial Neural NetworkNetwork. . Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 69-78

1. INTRODUCTION Human factor is often described as the weakest part of a i. What you know or knowledge-based systems: security system and users are often identified to be the a concept which has traditionally been weakest link in the security chain. Likewise, embodied in Personal Identification Numbers authentication is a cornerstone of most security systems (PINs) and passwords. today, and most users interact with these mechanisms on ii. What you have or token-based systems: a a daily basis. Some mechanisms can operate as a one- concept commonly related to smartcards. step procedure of identification or verification only. iii. Who you are or systems based on biometrics: Authentication mechanisms have been described to be of the notion related to biometric authentication. three types, namely:

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

These traditional authentication techniques have been One part can be stored with security environment used to provide a secure means to keep information whereas other part can embed in to any user-friendly secure. However, as our information rich society device. According to [7] in her paper, she defined the becomes heavily reliant on greater measures of security term user-centered security to refer to “security models, to protect critical information and data, it has proved mechanisms, systems, and software that have usability as extremely demanding to consider the issue of usability a primary motivation or goal.”She also emphasized the within the different authentication approaches. Usability importance of focusing on the users in the development can be defined as "the extent to which a product can be of secure systems. This is because when user’s mental used by specified users to achieve specified goals image of his protection goals matches the mechanisms he with effectiveness, efficiency and satisfaction in a must use, mistakes will be minimized. She also identified specified context of use"( International Standard three basic challenges facing user centered security as Organization (ISO). Security usability is concerned with well as identifying ways in which these problems are the study of how securing information should be handled being circumvented. The problems include: in relationship with the different human users. According to [3], authentication methods are needed and since (1) human and social relationships to usable security, security mechanisms are conceived, implemented, put (2) technical challenges best attacked with research, and into practice and violated by people, human factors (3) further difficulties with implementation and should be taken into account in their design. deployment.

Usability of the authentication mechanisms has seldom These troubles can be resolved by Human Computer been investigated as a result little has been said about Interaction techniques and some usable security suitable evaluation model that considers usability and principles to take us to the next level. According to her security. The aim of this research is to develop a model research expert evaluation and user testing are producing for evaluating a fully functional usable authentication effective usable security today. Principles such as safe method using artificial neural network due to its pressing staging, enumerating usability failure risks, integrated urgency. security, transparent security and reliance on trustworthy authorities can also form the basis of improved systems. 2. RELATED WORKS According to [6], specific risk and complications [1] took an exception to the popular opinion that “the associated with password authentication were human factor is often described as the weakest part of a highlighted. He also proposed an authentication method security system and users are often described as the based on neural network using feed-forward architecture. weakest link in the security chain” by suggesting that, The specific risks he highlighted includes: technical rather than blaming users, we should understand the roles (brute force), discovery and social engineering. In the and demands placed on them by security systems. He laid brute force attack, two methods can be used, emphasis on traditional password authentication (a) Attempting passwords against the system, but this is highlighting various problems and proffering solutions to easily stopped with account lockouts. these problems. Some of the problems he discovered (b) An offline attack against the password hash file. were based on the fact that users tend to choose short This is processor intensive search through the entire and/or guessable passwords , users forget their password key-space, calculating and comparing hash passwords, users are often willing to tell their passwords values of potential passwords to the values in the stolen to strangers who asked for them. hash file. He however carried out research that shows that users are Password may also be compromised by discovery. Forms not to be blamed entirely for many of these problems of password discovery may vary and include interception because users are often the inheritors of system defects, of a script file, an exploit on another system, a Trojan poor designs, incorrect installations, faulty operation, and program capturing keystrokes, or the discovery of default bad management. Other problems supposedly caused by passwords associated with other system or programs. The users happen because users are often faced with primary defense against discovery is proper system remembering dozens of passwords or PINs, but human design rules that do not allow discovery of passwords memory is brief and easily confused. Also users are through scripts or default system accounts. Social supposed to create passwords that cannot be guessed. engineering represents an attempt by an intruder to elicit However, our memory systems are particularly weak at password and account information from a user. This remembering meaningless content. Some of the solutions attack is exogenous to the computer system in question, to the abovementioned problems according to [1] coming via phone, fax, email or causal contact. [6] used includes: Users should also be educated about their artificial neural network to design a very efficient, robust password choices; reduce the memory load placed on and simple security system having the intrusion detection users as it is well known that cued recall, where users are capability. He achieved this by splitting the encrypted prompted for the information they must remember, is information in two parts. more accurate than free recall.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

[5] developed a model for evaluating the human impact Most biometric techniques are based on something that that password authentication issues are having on the cannot be lost or forgotten. While the advantages of security of information systems. This was done by biometric authentication look very attractive, there are carrying out a survey. The survey helped to evaluate user also many problems with biometric authentication that practices in determining passwords as well as determine one should be aware of. Biometric systems still need to vulnerabilities produced through user actions. A large be improved in the terms of accuracy and speed. Also not federal agency (US government agency) was used as the all users can use any given biometric system. People case study. The survey was followed by testing the without hands cannot use finger-print or hand- based usefulness of individuals customizing their passwords systems. Visually impaired people have difficulties using utilizing meaningful data and mnemonic devices in iris or retina based technique s. As not all users are able password development and also determining the human to use a certain biometric system, the authentication impact that password authentication issues have on system must be extended to handle users falling into the information security. The findings indicate that human “Failure To Enroll” category. This can make the resulting error associated with password authentication can be system more complicated, less secure or more expensive. significantly reduced through the use of passwords Biometric data are not considered to be secret and comprised of meaningful data for the user and that meet security of a bio-metric system cannot be based on the the information technology community requirement for secrecy of user’s biometric characteristics. However strength of password. according to the writer, sometimes biometric authentication systems replace traditional authentication [10] frowned at the use of alphanumeric password form systems not because of higher security but because of of authentication because of its inherent disadvantages. higher comfort and ease of use. The research also help to deduce that graphical passwords (i.e., passwords that are based on images [12] proposed a new graphical password scheme using rather than alphanumeric strings.) which has led to dynamic block-style that aims to balance the problem greater memorability, decrease in the tendency to choose between usability and security. Fuzzy logic methods were insecure passwords and increase in overall password used to enhance the usability by allowing certain degree security also has some disadvantages. These include of tolerance during authentication. The proposed scheme needing simple, artificial images, predefined regions, and is able to provide larger password space and reduces consequently many clicks in a password. In the paper a registration and authentication time. The results of the new design was explained which improved on the experiment was, in terms of usability, the registration and previous approaches in the areas of security, learning, authentication processes are simple; users only need to performance, and retention. In the proposed system any key in their username and click on the password image could be used and it does not need artificial (image).Selection of the image also affects the security predefined click regions with well-marked boundaries – a and usability of the system. Simple images increase the password can be any arbitrarily chosen sequence of usability, which are more users friendly and memorable. points in the image. From the results from various users The users did not get confused during the system test who tested the system some of the advantages of the phase. Besides, these processes are easy and fast. Users system were: It is easy to obtain large passwords spaces; are allowed to proceed to their actual tasks with minimal furthermore, in the experiment it appears that users rarely time. A degree of tolerance is allowed to increase the chose points that were within the tolerance around the system usability. Furthermore, the user interface of the click point of another participant. That is, people were system is simple and understandable to users on the first not strongly drawn to a few salient areas that an attacker view. Lastly, the instructions are clear to guide new users might guess. Finally, there is currently no efficient way on how to use the system. of creating dictionary attacks against the system. [9] explained the positive relationship between security [11] Summarizes the advantages and disadvantages of and usability. According to the paper, security is aimed at biometric authentication in terms of its usability and making undesirable actions more difficult while usability security. According to the writers, biometric aims at making desirable ones easier for the user, it may authentication methods are better than traditional also be true that improving one also improves the other. methods since these methods are solely based on A usable system will minimize unintentional errors, properties that can be forgotten, disclosed, lost or stolen. while a secure system will aim at ensuring that Biometrics was defined as automated methods of identity undesirable actions in a system are prevented or verification or identification based on the principle of mitigated. A security usability threat model was also measurable physiological or behavioral characteristics developed that focused on legitimate users’ mistakes that such as a finger-print, an iris pattern or a voice sample. It may compromise the system as against standard security also sheds light on the pros of biometric authentication. threat models which focus primarily on malicious They authenticate the user. Users cannot pass their attackers who may or may not be legitimate users. biometric characteristics to other users as easily as they do with their cards or passwords.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The concept of usage scenarios (usage scenarios are actions that are desirable to stakeholders of a secure system) and threat (negative) scenarios (threat scenarios are actions that are not desirable and hence the system should not allow them to happen) were used to understand and identify both system and external elements that are threats to a system’s usability, security, or both.

[2] summarized current research on the usability of security mechanisms and discussed options for increasing this usability and the effectiveness of these mechanisms. According to [2] usable security is not simply an issue of ‘fixing’ user interfaces to current mechanisms; rather, a change in how individuals, organizations and governments think about security is required. Effective security has to take into account the needs of all stakeholders, acknowledge that their needs sometimes Fig. 1: A Biological Neuron Source: [4] conflict and find a solution that is acceptable for all stakeholders in ongoing use. However while the paper The advantages of neural networks were said to be: emphasizes the urgent need to put users’ needs and i. Neural networks are data-driven self-adaptive values at the centre of security design, there is a note of methods in that they can adjust themselves to warning: most users are not knowledgeable about the data without any explicit specification of security, nor do they want to be. Motivational approaches functional or distributional form for the therefore could be employed to change underlying underlying model. perceptions about security and a limited set of key ii. Neural networks are universal functional behaviors, but they will not motivate the majority of approximators in that they can approximate users to become security experts. any function with arbitrary accuracy. Since any classification procedure seeks a functional [3] presented the usability and security issues of the user relationship between the group membership authentication methods in the computer security and and the attributes of the object, accurate access control domains. The various methods of identification of this underlying function is authentication were compared in terms of their security doubtlessly important. and usability. Some of the methods used includes iii. Neural networks are nonlinear models, which passwords, pins, proximity card, one time generators, makes them flexible in modeling real world challenge response, multifunction card, finger print, voice complex relationships. and signature. According to the survey carried out by the iv. Neural networks are able to estimate the writers, one of the methods that scored highest is the posterior probabilities, which provide the basis multifunction card. It however has the need of a smart for establishing classification rules and for card reader as its disadvantage. performing statistical analysis.

3. NEURAL NETWORKS Artificial neural networks (ANN) are relatively crude [3] and [8] explained that neural networks have emerged electronic networks of "neurons" based on the neural as an important tool for classification purposes. The structure of the brain. They process records one at a recent research activities in neural classification have time, and "learn" by comparing their classification of the established that neural networks are a promising record (which, at the outset, is largely arbitrary) with the alternative to various conventional classification known actual classification of the record. The errors methods. from the initial classification of the first record is fed back into the network, and used to modify the networks algorithm the second time around, and so on for many iterations. ANNs combine artificial neurons in order to process

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

information. A neuron in an artificial neural network is; 1. A set of input values (xi) and associated weights (wi), 2. A function (g) that sums the weights and maps the results to an output (y) [8].

4. USABILITY-CENTERED AUTHENTICATION

A usability-centered authentication approach is urgently needed to be able to bridge the gap between usability and security thus satisfying both usability and security goals of the users of authentication systems as describe in the figure below: Fig. 2: An Artificial Neuron Source: [4]

Fig. 3: A Trust Model For A Usability-Centered Authentication Approach

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

5. THE PROPOSED ARTIFICIAL NEURAL There are different methods of training, this includes NETWORK MODEL pattern association or pattern classification, but for the purpose of this research, the network will be trained for The basic network used for classifying the usability regression. The process of training usability information information is a feed-forward back propagation network. will then require a set of examples of proper network The arguments used for categorizing the usability behavior for which the default performance function for information are: feed-forward networks is mean square error.  Input vectors  Target vectors The network will then apply performance function to  Size of the layer employed in the neural network determine how to adjust the weights to minimize  Transfer function performance. The gradient will then be determined using  Backpropagation network training function a technique called back-propagation, which involves  Backpropagation weight/bias learning function performing computations backward through the network. The back-propagation computation is derived using the The first step in creating a feed forward network for chain rule of calculus in most of the training algorithms usability information is creating a network object this is discussed up to this point; a learning rate will then be achieved by using the new function command and its used to determine the length of the weight update (step argument are as listed above. The network is then used in size). The proposed Neural Network Architecture for the training the usability information the network weights evaluation of the usability of the different authentication and biases are initialized. methods is described in the figure below:

Input layer hidden layer 1 & 2

Fig. 4: Proposed Neural Network Architecture Usability-Centered Authentication Evaluation

For this research, a feed forward network is used because In a multilayer network based on the nonlinear sigmoid of its ability with non-linear classification problems. It is discriminant function. In a multilayer network the able to form from the high order morphological-extracted number of nodes is determined by the dimension of the features of Authentication factors. Both the hidden and feature space physically the hidden layers are output layers use a continuous network based on the non- inaccessible while the output layer provides user with linear sigmoid discriminant function. learning responses after training by adding more hidden layers the network is able to extract higher order status in order to perform more complex task.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Flow Chart The flow chart of the artificial neural network is described in figure below.

Input training START data set

Initialize weight randomly

Start training

present training set to compute outputs for all layers

Calculate cumulative error value NO

Compute error vectors and adjust weights accordingly

Are all pattern in one epoch represented ?

YES

Check if training completed and if terminal value is exceeded for training session

Is terminal value Is training YES exceeded for NO completed? NO training Session ?

YES

Output Save weights epochs and errors STOP Result

Fig. 5 Flow Diagram for The Application Process

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The basic model we use for the study is logistic 6. CONCLUSION AND FUTURE WORKS regression because the responses are in categorical form. Logistic regression analyzes binomially distributed data In this work, the Neural Network based model for of the form. evaluating a fully functional usable Authentication method has been developed to ascertain Y ~ B(n , p ), for i 1, . . . , n usability in the authentication process, thereby providing i i i a suitable evaluation model that considers usability and security. This is because over the years it has proved where the numbers of trials ni are known and the extremely demanding to merge usability with security in probabilities of success pi are unknown. The model the choice of authentication methods. proposes that for each trial there is a set of explanatory In view of the proposed Neural Network based model for variables which can be thought of as being in vector Xi evaluating the usability of the different authentication and the model therefore takes the form: methods, we hope to collect the opinions of a range of  Y  computer users through questionnaires as to which i authentication approach they find most desirable. The pi  E X i    results captured will be used to train the artificial neural  ni  network and the neural network model will also be able

to predict the authentication approach that achieves both The logits of the unknown binomial probabilities are usability and security preferences of various users. modeled as a linear function of the X , that is, i

 p  z  logit( p )  ln i      x   x   x  . . .   x i   0 1 1 2 2 3 3 k k 1 pi 

We use Logistic regression to predict the probability of occurrence of an event by fitting data to a logistic curve. In logistic regression model, the relationship between ‘input’, z, and the probability of the event of interest is described by the function. 1 f (z)  1 e z The variable z is known as the logit and is usually defined as z   0  1 x1   2 x2   3 x3  . . .   k xk

Therefore, it follows that

1 f (z)  (0 1x1 2 x2 3x3  . . . k xk ) 1 e The model also assumes that z is linearly related to the predictors.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFERENCES Author’a Biography 1. Andrew Patrick, 2002:”Human Factors of Security Systems: A Brief Review” 2. Angela Sasse, M., 2004:”Usability and trust in Mapayi T. is a doctoral information systems”. research student at the Christina Bra and Jean-Marc Robert, 2006:” Department of Computer Security and Usability: The Case of the User Science, University of KwaZulu- Authentication Methods”. Natal, Durban, South Africa. He 3. Dave Anderson and George McNeill, 1992: ” had his BSc in Computer ARTIFICIAL NEURAL NETWORKS Science at the University of TECHNOLOGY A DACS State-of-the-Art Ado-Ekiti, and his Msc at the Report”, August 20. University of Ibadan, Nigeria. He is a member of 4. Deborah Sater Carstens, Pamela R. McCauley- Computer Professionals of Nigeria. His areas of research Bell, Linda C. Malone and Ronald F. DeMara, includes; Biometrics, Computer Security, Computer 2004: “Evaluation of the Human Impact of Vision and Pattern Recognition and ICT for Password Authentication Practices on Development,. He can be reached through Information Security”. Email: [email protected]. 5. Manoj Kumar Singh, 2009:”Password based: A generalized robust system design using neural network”. 6. Mary Ellen Zurko, 2011:”User-Centered Security: stepping Up to the Grand Challenge”. 7. Meral Yay and Eylem Deniz Akıncı, Engr. Olaniyan Olatayo 2009:”Application of Ordinal Logistic Moses (M.SC), a Lecturer at Regression and Artifical Neural Networks in a Bells University of Study of Student Satisfaction”. Technology, Ota, Nigeria. He 8. Ronald Kaind, Ivan Flechais and RoscoeA.W., teaches computing in the Dept 2008:“Security and Usability: Analysis and of Computer Science and Evaluation”. Information Technology. Engr 9. Susan Wiedenbeck, Jim Waters, Jean-Camille Olaniyan obtained the B.Tech Birget, Alex Brodskiy and Nasir Memon, Computer Engineering degree from the Ladoke Akintola 2005:” Authentication Using Graphical University of Technology, Ogbomoso and a MSc in Passwords: Basic Results”. Computer Science from the University of Ibadan, Ibadan, 10. V´aclav Maty´aˇs and Zdenˇek ˇR´ıha, 2005: Nigeria. Presently on PhD degree focusing on “Biometric Authentication —Security and Telemedicine Research his other research interest Usability” includes Computer Networks and Human Computer 11. Woo Chaw Seng, Yong Kok Khuen and Ng Interaction. He can be reached through E-mail: Liang Shing, 2011:”Enhanced Graphical [email protected] Password by using Dynamic Block-style Scheme”.

Shakirat. O.Raji is a PhD Research Student at Human- Centered Design Group ( HCDG), Department of Information Systems, Faculty of Information and Communication Technology International Islamic University Malaysia. Her Main area of Interest is Human-Computer- Interaction and part of her interests are Usability Evaluation, Usability Engineering, user Experience as well as Cyber Security. She can be reached through [email protected]

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Olaifa M. had a B.Sc Computer science from the University of Ilorin and his MSc in Computer science from the University of Ibadan. He was a lecturer at Houdegbe North American University, Benin, Republic of Benin between 2009 and 2011. He is currently at The Department of Computer systems engineering, school of Information and communication technology, Tshwane University of Technology, Pretoria, South Africa. His research area includes Machine Learning (intelligent industrial systems).He can be reached through E-mail: [email protected]

Isamotu Nehemiah O. is a graduate of Bells University of Technology, Ota, Nigeria. He can be reached through E-mail: [email protected]

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

On Modeling Confidentiality Archetype and Data Mining in Cloud Computing

Alawode A. Olaide Dept of Computer Systems Engineering. Imo State University Oweri, Imo State, Nigeria [email protected]

ABSTRACT Cloud computing is an Internet-based computing, whereby shared resources, software, and information are provided to computers and other devices on demand. It is an upcoming paradigm that offers tremendous advantages in economical aspects, such as reduced time to market, flexible computing capabilities, and limitless computing power. Popularity of cloud computing is increasing day by day in distributed computing environment. There is a growing trend of using cloud environments for storage and data processing needs. To use the full potential of cloud computing, data is transferred, processed, retrieved and stored by external cloud providers. However, data owners are very skeptical to place their data outside their own control sphere. Their main concerns are the confidentiality, integrity, security and methods of mining the data from the cloud. This paper discus efforts directed to which degree this skepticism is justified, by proposing to model Cloud Computing Confidentiality Archetype and Data Mining 3CADM. The 3CADM is a step-by-step framework that creates mapping from data sensitivity onto the most suitable cloud computing architecture and process very large datasets over commodity clusters with the use of right programming model. To achieve this, the 3CADM determines the security mechanisms required for each data sensitivity level, which of these security controls may not be supported in certain computing environments, which solutions can be used to cope with the identified security limitations of cloud computing. The model achieves data confidentiality while still keeping the harmonizing relations intact in the cloud. It also achieves an algorithm to mine the data from the cloud using sector/sphere framework with association rules.

Keywords: Cloud, Computing, Archetype, Data Mining, Confidentiality, Modelling.

African Journal of Computing & ICT Reference Format Alawode, A.O. (2013). On Modeling Confidentiality Archetype and Data Mining in Cloud Computing. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 79-86.

1.0 Introduction Cloud computing is the delivery Cloud computing can be the ability to use applications on of computing and storage capacity as a service to a the Internet that store and protect data while providing a community of end-recipients. The name comes from the service — anything including email, sales force use of a cloud-shaped symbol as an abstraction for the automation and tax preparation. It can be using a storage complex infrastructure it contains in system diagrams. cloud to hold application, business, and personal data. Cloud computing entrusts services with a user's data, And it can be the ability to use a handful of Web services software and computation over a network. There are to integrate photos, maps, and GPS information to create various opinions on what is cloud computing. It can be a mashup in customer Web browsers. the ability to rent a server or a thousand servers and run a geophysical modeling application on the most powerful Sun’s MicroSystems in a white paper titled Introduction systems available anywhere. It can be the ability to rent a to Cloud Computing architecture, 2009 takes an virtual server, load software on it, turn it on and off at inclusive view that there are many different types of will, or clone it ten times to meet a sudden workload clouds, and many different applications that can be built demand. It can be storing and securing immense amounts using them. To the extent that cloud computing helps to of data that is accessible only by authorized applications increase the velocity at which applications are deployed, and users. It can be supported by a cloud provider that helping to increase the pace of innovation, cloud sets up a platform that includes the OS, Apache, a computing may yet take forms that we still cannot MySQL™ database, and PHP with the ability to scale imagine today. What distinguishes cloud computing from automatically in response to changing workloads. previous models? Boiled down to a phrase, it’s using information technology as a service over the network. Sun define it as services that are encapsulated, have an API, and are available over the network.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Fig 1: Cloud computing logical diagram Source: en.Wikipedia.org

This definition encompasses using both computer and “Cloud computing is a model for enabling storage resources as services. Cloud computing is based convenient, on-demand network access to a on the principle of efficiency above all — efficiency that shared pool of configurable computing produces high-level tools for handling 80% of use cases resources (e.g., networks, servers, storage, so that applications can be created and deployed at an applications, services) that can be rapidly astonishing rate. Cloud computing can be provided using provisioned and released with minimal an enterprise datacenter’s own servers, or it can be management effort or service provider provided by a cloud provider that takes all of the capital interaction” risk of owning the infrastructure. The illusion is that . resources are infinite. While the field is in its infancy, the To explain the definition in short, “convenient on- model is taking the information technology (IT) world by demand network access”, together with “minimal storm. management effort or service provider interaction,” stands for easy and fast network access to resources that The U.S. National Institute of Standards and Technology are ready to use. With a “shared pool of resources,” the (NIST) have put an effort in defining cloud computing. available computing resources of a cloud provider are Since NIST‟s publications are generally accepted, their combined as one big collection, to serve all users. The definition of cloud computing will be used in this “rapid provisioning and releasing” of computing dissertation. The NIST definition of cloud computing is resources is used to quickly match available resources, (NIST 2009a) with the need for those resources. This rapid provisioning prevents a lack of computing power when the need increases, while rapid release of assigned resources prevents that resources are idle while they may be required elsewhere.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The above definition is by no means exhaustive and it is  Which data classifications are used today and very hard to find two experts having the same definition what are their security requirements with of cloud computing. Cloud computing is still an evolving respect to security and confidentiality? paradigm. The characteristics, deployment and delivery models, as well as the underlying risks, technologies,  Which cloud architectures are available and issues and benefits will be refined by energetic debate by what security controls do they have in place both the public and the private sectors. A more elaborate with respect to confidentiality? explanation of these cloud properties will be discussed in this dissertation. As with most new technologies and  How can we classify cloud architectures in the paradigms, users usually look for the functionality first area of confidentiality and security? and only later on, looks after the security and integrity of such functionality. Cloud computing also raises such an  How can we create a mapping from amount of questions concerning security guarantees that confidential data classes to cloud architectures potential users are waiting for clear answers before operating on potentially confidential data? moving into the cloud.  What are the data mining parameters as it 2. RESEARCH MOTIVATION AND OBJECTIVES affects cloud computing architectures?

Cloud computing allows large organizations to tap into a virtually infinite pool of resources with the ability to 4. RESEARCH SCOPE control cost. Some cloud providers give this power to their customers by offering a pay-as-you-go pricing A broad approach of for this research is to classify cloud model as well as elasticity and availability of computing assets and networks on the topic of confidentiality and and storage resources. Cloud computing users work with security. Investigating various security and data mining data, software and applications that are located outside models is highly essential. In this dissertation, we shall their premise. Due to this benefit, cloud computing has focus on cloud’s data security, confidentiality and quickly gained popularity in recent years. Yet many mining, as these are where the major concerns are at this individuals and organizations have not widely adapted moment. the use of clouds due to the concerns of confidentiality, security, and privacy of their data and the way they are Data classification research has already been done mined. Many organizations are uncomfortable with the extensively (Chen and Liu 2005; Morsi, El-fouly and idea of having their data, software and applications on Badr 2006; Grandison, Bilger, O'Connor et al. 2007), this systems they do not control. There is a lack of knowledge dissertation will use the results of these researches and on how cloud computing impacts the confidentiality of analyze the security requirements that need to be met in data stored, processed and transmitted in cloud order to protect data confidentiality. computing environments.

The goal of this dissertation is to create a model with the 5. RESEARCH SIGNIFICANCE use of algorithm programming tools to resolve these issues and make public clouds more secure and easily Presenting a model for cloud computing confidentiality mined. The model clarifies the impact of cloud archetype and data mining while still keeping the computing on confidentiality preservation, by making harmonizing relations intact in the cloud will stepwise recommendations on how data can be classified significantly resolve the concerns of cloud computing on confidentiality, how data classifications relate to the users on the security and confidentiality of their data security controls need to preserve the confidentiality of stored and accessed outside their locations. data, how the process of security control selection is negatively influenced in cloud computing environments and how to cope with the negative influences of cloud 6. RESEARCH LIMITATION computing on the protection of data confidentiality. Attending and participating in both local and international seminars, Research forums, conferences and 3. RESEARCH QUESTIONS workshops are very essential in carrying out this research project. However, cost is a major threat. In order to achieve the research objectives stated above, the necessary knowledge will need to be obtained and combined. The following research questions will guide this research:

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

7. RESEARCH PLAN AND METHODOLOGY 8.1 Cloud key characteristics

The research starts with the orientation on the area of 8.1.1 On-demand self-service. Cloud computing cloud computing, what is cloud computing about and resources can be procured and disposed of by the which security issues are in dire need of investigation. By consumer without human interaction with the cloud consulting websites of current cloud service offerings, service provider. This automated process reduces the reading news articles, participating in seminars and personnel overhead of the cloud provider, cutting costs discussing cloud computing and security issues with and lowering the price at which the services can be professionals within and outside Nigeria. The research offered. methodology employed in this project is structured system analysis and design methodology with its object 8.1.2 Resource pooling. By using a technique called oriented partners. (SSADM/OOADM). “virtualization,” the cloud provider pools his computing resources. This resource pool enables the sharing of Structured Systems Analysis and Design Methodology is virtual and physical resources by multiple consumers, a systems approach to the analysis and design of “dynamically assigning and releasing resources according information systems. This approach is mainly designed to consumer demand” (NIST 2009a). The consumer has for large scale information systems with high volume of no explicit knowledge of the physical location of the business events. SSADM is a waterfall method by which resources being used, except when the consumer requests an information system can be arrived at. Object oriented to limit the physical location of his data to meet legal modeling and design promote better understanding of requirements. requirements, cleaner designs, and more maintainable systems. An object oriented systems analysis and 8.1.3 Broad network access. Cloud services are designs can be used to analyse problems requirements, accessible over the network via standardized interfaces, design a solution to the problem, and implement a enabling access to the service not only by complex solution in a programming language or database. devices such as personal computers, but also by light (Alawode, 2009) weight devices such as smart phones.

8.1.4 Rapid elasticity. The available cloud computing 8. BACKGROUND OF STUDY resources are rapidly matched to the actual demand, quickly increasing the cloud capabilities for a service if

the demand rises, and quickly releasing the capabilities As the paradigm of cloud computing is relatively new, when the need for drops. This automated process there are various open issues which need to be resolved decreases the procurement time for new computing before cloud computing is fully accepted by the broad capabilities when the need is there, while preventing an community. A deeper explanation is needed of what abundance of unused computing power when the need cloud computing encompasses. has subsided.

The NIST definition of cloud computing mentioned in 8.1.5 Measured service. Cloud computing enables the the introduction will be used as our starting point. To measuring of used resources, as is the case in utility recall the definition: computing. The measurements can be used to provide

resource efficiency information to the cloud provider, and “Cloud computing is a model for enabling can be used to provide the consumer a payment model convenient, on-demand network access to a based on “pay-per-use.” For example, the consumer may shared pool of configurable computing be billed for the data transfer volumes, the number of resources (e.g., networks, servers, storage, hours a service is running, or the volume of the data applications, services) that can be rapidly stored per month. provisioned and released with minimal

management effort or service provider 8.2 Cloud Service Models interaction” 8.2.1 Software-as-a-Service (SaaS). The SaaS service

model offers the services as applications to the consumer, The above definition is supported by five key cloud using standardized interfaces. The services run on top of characteristics, three delivery models and four a cloud infrastructure, which is invisible for the deployment models (NIST 2009a). These supporting consumer. The cloud provider is responsible for the properties will be explained and we will discuss various management of the application, operating systems and security issues and concerns related to cloud computing. underlying infrastructure. The consumer can only control

some of the user-specific application configuration

settings.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

8.2.2 Platform-as-a-Service (PaaS). The PaaS service users are also considered as trusted by the organizations model offers the services as operation and development that are part of the community. platforms to the consumer. The consumer can use the platform to develop and run his own applications, 8.3.4 Hybrid clouds. Hybrid clouds are a combination of supported by a cloud-based infrastructure. “The public, private, and community clouds. Hybrid clouds consumer does not manage or control the underlying leverage the capabilities of each cloud deployment cloud infrastructure including network, servers, operating model. Each part of a hybrid cloud is connected to the systems, or storage, but has control over the deployed other by a gateway, controlling the applications and data applications and possibly application hosting that flow from each part to the other. Where private and environment configurations” (NIST 2009a). community clouds are managed, owned, and located on either organization or third party provider side per 8.2.3 Infrastructure-as-a-Service (IaaS). The IaaS characteristic, hybrid clouds have these characteristics on service model is the lowest service model in the both organization and third party provider side. The users technology stack, offering infrastructure resources as a of hybrid clouds can be considered as trusted and service, such as raw data storage, processing power and untrusted. Untrusted users are prevented to access the network capacity. The consumer can use IaaS based resources of the private and community parts of the service offerings to deploy his own operating systems hybrid cloud. and applications, offering a wider variety of deployment possibilities for a consumer than the PaaS and SaaS 8.4 Cloud Security Issues models. “The consumer does not manage or control the Although it is important to describe the location of the underlying cloud infrastructure but has control over security perimeter in relation to the assets to be protected, operating systems, storage, deployed applications, and using the terms external clouds and internal clouds would possibly limited control of select networking components indicate a well-defined perimeter between the outside and (e.g., host firewalls)” (NIST 2009a). the protected inside. This separation is an anachronistic concept due to the de-perimeterization and the loss of 8.3 Cloud deployment models trust boundaries resulting from the increasing need of Regardless of which delivery model is utilized, cloud companies to collaborate and provide ubiquitous access offerings can be deployed in four primary ways, each to employees, consumers and contractors. with their own characteristics. The characteristics to describe the deployment models are; (i) who owns the Traditional security controls may be incapable to handle infrastructure; (ii) who manages the infrastructure; (iii) the shift from secure silos of data with strict trust where is the infrastructure located; (iv) and who accesses boundaries and well defined access control, to the the cloud services. complex scenarios where access is ubiquitous, information exchange is abundant and data location is 8.3.1 Public clouds. Public cloud computing is based on often unknown. Cloud computing accelerates this erosion massive scale offerings to the general public. The of trust and security boundaries. infrastructure is located on the premises of the provider, who also owns and manages the cloud infrastructure. With cloud computing, organizations can use services Public cloud users are considered to be untrusted, which and store data outside their own control. This means they are not tied to the organization as employees development raises security questions and should induce and that the user has no contractual agreements with the a degree of skepticism before using cloud services. In his provider. article, Brodkin discusses a study of Gartner, which points out seven areas of concern around security issues 8.3.2 Private clouds. Private clouds run in service of a in cloud computing (Brodkin 2008): single organization, where resources are not shared by other entities. “The physical infrastructure may be owned 8.4.1 Privileged user access by and/or physically located in the organization‟s Data stored and processed outside the enterprises direct datacenters (on-premise) or that of a designated service control, brings with “an inherent level of risk, because provider (off-premise) with an extension of management outsourced services bypass the physical, logical and and security control planes controlled by the organization personnel controls IT shops exert over in-house or designated service provider respectively“ (Bardin, programs” (Brodkin 2008). Brodkin advises to get as Callas, Chaput et al. 2009). Private cloud users are much information as possible about the people who considered as trusted by the organization, in which they manage your data and the controls they implement. are either employees, or have contractual agreements with the organization. 8.4.2 Regulatory compliance Data owners are responsible for the integrity and 8.3.3 Community clouds. Community clouds run in confidentiality of their data, even when the data is outside service of a community of organizations, having the same their direct control, which is the case with external deployment characteristics as private clouds. Community service providers such as cloud providers. Where

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

traditional service providers are forced to comply to a contractual statement specifying support for external audits and obtain security certifications, so incorruptible logging and investigation, Gartner says that should cloud computing providers: “Cloud computing “the only safe assumption is that investigation and providers who refuse to undergo this scrutiny are discovery requests will be impossible” (Gartner 2008). signaling that customers can only use them for the most trivial functions” (Brodkin 2008). Most, if not all, of the 8.4.7 Data Lock-in leading cloud providers do not support on-site external Availability of customers data may be at risk if a cloud audits on customers request. As a result, some provider goes broke or is acquired by another compliances cannot be achieved because on-site auditing organization. Providers should provide procedures how is a requirement that cannot be satisfied, for example the customers can retrieve their data when the needed, and at Payment Card Industry level 1 compliancy. least as important; in which format the data is presented to the customer. If the data is presented in a format 8.4.3 Data location proprietary to the cloud provider, it may be unusable by The exact location of data in the cloud is often unknown. any other provider. The use of open standards by Data may be located in systems in other countries, which providers to prevent data lock-in is recommended, but not may be in conflict with regulations prohibiting data to always supported. leave a country or union. Gartner advises to investigate if cloud providers will commit to keeping data in specific Of the above security issues, the issues related to jurisdictions and whether the providers will make availability of services are well attended to by researchers contractual commitments to obey local privacy and cloud service providers. The largest uncertainties requirements on behalf of their customers (as cited in linger around issues related to confidentiality of data, Brodkin, 2009). For example, the EU Data Protection such as data location, access control and regulatory Directive places restrictions on the export of personal compliance. data from the EU to countries whose data protection laws are not judges as “adequate” by EU standards (European Commission 1995a). If not properly attended to, 8.5 Issues of Data Mining in the Cloud European personal data may be located outside the EU without being compliant to the directive. Data mining, the extraction of hidden predictive 8.4.4 Data Segregation information from large databases, is a powerful new The shared, massive scale characteristics of cloud technology with great potential to help companies focus computing makes it likely that one‟s data is stored on the most important information in their data alongside data of others consumers. Encryption is often warehouses. (Bhagyashree, et al 2012). used to segregate data-at-rest, but it is not a cure-all. It is advised to do a thorough evaluation of the encryption Data mining tools predict future trends and behaviors, systems used by the cloud provider. A proper built, but allowing businesses to make proactive, knowledge-driven poorly managed encryption scheme may be just as decisions. The automated, prospective analyses offered devastating as no encryption at all, because although the by data mining move beyond the analyses of past events confidentiality of data may be preserved, availability of provided by retrospective tools typical of decision data may be at risk when data availability is not support systems. As data sets have grown in size and guaranteed. complexity, direct hands-on data analysis has increasingly been augmented with indirect, automatic data processing. This has been aided by other discoveries 8.4.5 Recovery in computer science, such as neural networks, cluster Cloud providers should have recovery mechanisms in analysis, genetic algorithms (1950s), decision trees place in case of a disaster. “Any offering that does not (1960s) and support vector machines (1990s). Data replicate the data and application infrastructure across mining is the process of applying these methods to data multiple sites is vulnerable to a total failure,” Gartner with the intention of uncovering hidden patterns in large says (as cited in Brodkin, 2009). Cloud providers should data sets. Data mining is sorting through data to identify provide its guidelines concerning business continuity patterns and establish relationships. planning, detailing how long it will take for services to be fully restored. 8.5.1 Data mining parameters include:

8.4.6 Investigative support 1. Association - Looking for patterns where one event is Gartner warns that “investigating inappropriate or illegal connected to another event. activity may be impossible in cloud computing, because 2. Sequence or path analysis - Looking for patterns where logging and data may be co-located and spread across one event leads to another later event ever-changing sets of hosts and data centers” (Brodkin 3. Classification - Looking for new patterns 2008). If cloud providers cannot provide customers with

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4. Clustering - Finding and visually documenting groups for their users. Here we explore the how the data mining of facts not previously known tools like SAS, PAS and IaaS are used in cloud 5. Forecasting - Discovering patterns in data that can lead computing to extract the information. A cloud provider to reasonable predictions about the future. This area for a data mining and natural language processing of data mining is known as predictive analytics. system. Leading cloud computing providers Amazon Web Services, Windows Azure, OpenStack. People use Example, the Visual Numeric’s has been providing this feature to build information listing, get information advanced forecasting and data mining solutions across a about different topics by searching in forums etc. wide range of industries such as aerospace, government, Companies use this service to see what kind of telecommunications, financial services and healthcare. information is floating in the world wide web for their Visual Numeric’s’ forecasting solutions combine products or services and take actions based on the data technical expertise, decades of hands-on experience and presented The information retrieval practical model powerful products to create the highest quality solutions through the multi-agent system with data mining in a possible for your visual data analysis needs. Same as, so cloud computing environment has been proposed. It is there are various applications of data mining in real however, recommended that users should ensure that the world. So there are many applications of Data mining in request made to the IaaS is within the scope of integrated real world As, Hospital, Student Management, Airline data warehouse and is clear and simple. Thus, making the Reservation, Forecasting, Biometrics, Mathematics, work for the multi-agent system easier through Geographical, Web Mining, Parallel Processing, Space application of the data mining algorithms to retrieve Organization, Data Integrity, etc. So there are many meaningful information from the data warehouse. Cloud application in which the data mining term is very useful. computing allows the users to retrieve meaningful information from virtually integrated data warehouse that

reduces the costs of infrastructure and storage. 8.5.2 Data mining in the Cloud

The Microsoft suite of cloud-based services includes a new technical preview of Data Mining in the Cloud “DMCloud”. DMCloud allows you to perform some 10.0 Proposed Contribution to ICT body of basic data mining tasks leveraging a cloud-based knowledge Analysis Services connection. (Bhagyashree, et al 2012) At the end of this dissertation, 3CADM (Cloud Computing Confidentiality Archetype and Data Mining) DMCloud is valuable capability for IWs that would like would be created. The goal is to present a model with the to begin considering SQL Server Data Mining without use of algorithm programming tools to resolve security, the added burden of needing a technology professional to confidentiality and data mining issues and make public first install Analysis Services. Additionally, IWs can use clouds more secure and easily mined. the DMCloud services no matter where they may physically be located as long as they have an Internet connection! The data mining tasks you can perform with The model will clarify the impact of cloud computing on DMCloud are the same Table Analysis Tools found in the confidentiality preservation, by making stepwise traditional Excel Data Mining add-in. These data mining recommendations on how data can be classified on tasks include: confidentiality, how data classifications relate to the security controls need to preserve the confidentiality of  Analyze Key Influencers data, how the process of security control selection is  Detect Categories negatively influenced in cloud computing environments  Fill From Example and how to cope with the negative influences of cloud  Forecast computing on the protection of data confidentiality.  Highlight Exceptions

 Scenario Analysis  Prediction Calculator  Shopping Basket Analysis

Data mining is used in various applications such as Health care, Student management, mathematics, Science, in various website. Data mining in cloud computing is the process of extracting structured information from unstructured or semi-structured web data sources. The data mining in Cloud Computing allows organizations to centralize the management of software and data storage, with assurance of efficient, reliable and secure services

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFERENCES NIST. (2004a). FIPS 199: Standards for Security Categorization of Federal Information and Alawode A.O. (2009). M.Sc. Thesis. A Computerized Information Systems. Retrieved August Network Based Security Information System 28, 2009, from for Nigerian Security Agencies http://csrc.nist.gov/publications/fips/fips199/FI PS- Bardin, J., Callas, J., Chaput, S., Fusco, P., Gilbert, F. et al. (2009). Security Guidance for Critical Areas NIST. (2004b). Guide for the Security Certification and of Focus in Cloud Computing v2.1, Retrieved Accreditation of Federal Information January 28, 2010, from Cloud Security Systems, SP 800-37. Retrieved January 30, Alliance, from 2010, from http://www.cloudsecurityalliance.org/guida http://csrc.nist.gov/publications/nistpubs/800- nce/ 37/SP800-37-final.pdf.

Bhagyashree A., Vaishali B. (2012). Data Mining in NIST. (2006). FIPS 200: Minimum Security Cloud Computing. “Recent Trends in Requirements for Federal Information and Computing” Proceedings published by Information Systems. Retrieved International Journal of Computer December 1, 2009, from Applications® (IJCA)ISSN: 0975 - 8887 http://csrc.nist.gov/publications/fips/fips200/FI PS- 200-final-march.pdf. Brodkin, J. (2008). Gartner: Seven cloud-computing security risks, Retrieved September 23, 2009, NIST. (2008a). Guide for Mapping Types of Information from Network World, from and Information Systems to Security http://www.networkworld.com/news/2008/070 Categories, Volume I, SP 800-60 Rev. 1. 208-cloud.html Retrieved August 27, 2009, from http://csrc.nist.gov/publications/nistpubs/800- Chen, K. and Liu, L. (2005). Privacy preserving data 60-rev1/SP800- 60_Vol1- Rev1.pdf. classification with rotation perturbation. In Proceedings of Fifth International Conference NIST. (2008c). Guide for Assessing the Security Controls of Data Mining, 589–592. IEEE. in Federal Information Systems, SP 800- 53A. Retrieved November 11, 2009, from

http://csrc.nist.gov/publications/PubsSPs.html. EuropeanCommission (1995a). 95/46/EC of the

European Parliament and of the Council of 24 NIST. (2009). National Institute of Standards and October 1995 on the protection of individuals Technology, main website. from with regard to the processing of personal data http://www.nist.gov. and on the free movement of such data. Official

Journal of the EC. 23. NIST. (2009a). NIST Working Definition of Cloud

Computing v15. Retrieved October 7, 2009, Gartner (2008). Assessing the Security Risks of Cloud from Computing, Retrieved December 5, 2009, http://csrc.nist.gov/groups/SNS/cloud- http://www.gartner.com/DisplayDocument?id= computing/index.html. 685308 NIST. (2009b, 12 August 2009). Recommended Security Grandison, T., Bilger, M., O'Connor, L., Graf, M., Controls for Federal Information Systems and Swimmer, M. et al. (2007). Elevating the Organizations, SP 800-53 Rev 3 from Discussion on Security Management: The Data http://csrc.nist.gov/publications/nistpubs/800- Centric Paradigm. In Proceedings of 2nd 53-Rev3/sp800-53-rev3-final-errata.pdf. IEEE/IFIP International Workshop, 84-93. Morsi, W., El-fouly, T. and Badr, A. (2006). Using IPSec NIST. (2001). Risk Management Guide for Information to Secure Multi-Level Data Classification in Technology Systems, SP 800-30 Retrieved MLS Networks. In Proceedings of 6th November 23, 2009, from International Conference on ITS http://csrc.nist.gov/publications/PubsSPs.html. Telecommunications, 817-821. Suns MicroSystems (2009). Introduction to Cloud Computing architecture White Paper 1st Edition, June 2009

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

An Experimental Comparison of Speech and DTMF for VoiceXML-Based Expert Systems

Oyelami Olufemi Moses & Akinyemi Ibidapo Olawole4 Department of Computer and Information Sciences, Covenant University, Ota, Nigeria [email protected], [email protected]

Uwadia Charles Onuwa Department of Computer Science, University of Lagos, Akoka, Lagos, Nigeria [email protected]

Akinwale Adio Taofeek3 Department of Computer Science, University of Agriculture, Abeokuta, Nigeria [email protected]

ABSTRACT Comparisons of DTMF and speech modalities for interacting with diverse dialogue systems for different tasks, among different user populations have led to different design recommendations for different user populations. This paper reports the results of the experimental comparison of these input modalities in a new context of VoiceXML-based diseases diagnosis expert system among a new user population - Nigerians. The results show that DTMF was more satisfying than speech for system satisfaction. Modality wise, speech was more satisfying than DTMF. Speech was also more natural than DTMF. DTMF was preferred by the majority and was more effective and efficient than speech. For diseases diagnosis expert health dialogue systems in Nigeria, DTMF is recommended for effectiveness and efficiency. It is also recommended for satisfaction. Speech is recommended for modality satisfaction while both modalities are recommended for entertainment purpose. Speech is advocated for modality naturalness. However, a platform that incorporates the two modalities will provide the benefits of the two, and allow the users varieties of choices that best suit their needs.

Keywords- DTMF; speech, expert system; VoiceXML; diagnosis.

African Journal of Computing & ICT Reference Format Oyelami Olufemi Mose, Uwadia Charles Onuwa, Akinwale Adio Taofeek & Akinyemi Ibidapo Olawole (2013). An Experimental Comparison of Speech and DTMF for VoiceXML-Based Expert Systems. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 87-94

1. INTRODUCTION It has also been found out that user preference for either VoiceXML as an extensible mark-up language was DTMF or speech depends on the nature of task performed designed for human-computer audio dialogues that [4]. In a previous research work, the authors have used feature synthesized speech, digitized audio, recognition component orientation approach to develop VoiceXML- of spoken input, Dual Tone Multi Frequency (DTMF) based expert system - Health Dialogue Expert System key input, recording of spoken input, telephony and (HDES) that acts as a physician to diagnose some mixed initiative conversations [1,2]. The language does selected types of fever rampant in Nigeria - malaria, not, however, support sophisticated dialogue systems that lassa, yellow and typhoid with a view to augment are artificial intelligence-based. In addition, users of healthcare [6] as the ratio of physicians to citizens in this VoiceXML-based systems have the options of either part of the world is abysmally low [7]. HDES interacting with them using speech or DTMF. Various incorporates Java expert system shell (Jess) into research work have compared the two input modalities in VoiceXML-based system as the expert system different contexts and among different user populations component using component orientation. In developing [3, 4, 5] and have come out with different results and HDES, five physicians were interviewed as to the design recommendations. There is a need to study the symptoms and how to diagnose the fever types handled characteristics of different user populations as relating to by HDES. Literature was also consulted. The same dialogue systems. One of the reasons for that is that, physicians tested HDES and approved its reliability. different user groups speak with different accents and exhibit different characteristics. As at 2007, Nigeria had an estimated population of 140 million, but the ratio of doctors to the population was

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

about 1 to 3,333 [8]. The density of physicians per 10, printed materials. “All the health contents and prompts 000 population was 4, that of dentists was .5, 16 for were translated using a registered translation service into nurses and midwives, and 1 for pharmaceutical the dialect of Setswana spoken in Botswana”. The study personnel. According to the World Health Organization was a within-subjects comparison, and after the (WHO) 2010 report, in Nigeria, current life expectancy is completion of each modality’s trials by the 33 caregivers 49 years. The most common causes of death in rank order involved, a post questionnaire was administered verbally are as follows: malaria, diarrhea, other diseases, to each of them. pneumonia, prematurity, birth asphyxia, neonatal sepsis, HIV/Aids, congenital abnormalities and injuries [7]. As The results obtained showed that there were no can be seen, malaria is actually the number one killer significant differences between task completion rates disease in Nigeria. There is therefore a need to fight the though speech performed slightly better that DTMF. 59% disease and other types of fever with all armories. of the users preferred DTMF while 19% preferred Currently, the government is encouraging the populace speech. This is in contrast to studies carried out in the through radio broadcasting to see their doctors for proper developed world where users preferred speech. The diagnosis instead of assuming they have malaria and self- results, however, correlate with the fact that simple medicate. DTMF is generally not viewed as favourable. The subjects that preferred speech did not as in the studies in However, the use of mobile phone in the country has the developed world comment that speech is more been on the rise. As at October 2012, the number of entertaining or enjoyable but rather on the utility of connected lines stood at 138, 029, 637 for mobile (GSM speech as being more accessible for older people or and CDMA) and fixed wired and wireless while the total faster” [3]. number of active lines stood at 109, 499,882 [9] out of the population of 140, 431, 790 [10]. This provides an Neil et al.[4] compared speech and dialed input voice avenue to reach a vast majority of the populace for user interfaces for farmers in Gujarat, India. They healthcare which has hitherto been inadequate. This “designed Avaaj Otalo (“voice-based community article reports the results of the experiments conducted to forum”), a Gujarati language application allowing compare the use of DTMF and speech as input modalities farmers to access agricultural information over the phone. for interacting with HDES in order for callers to diagnose Navigational nodes in the application were limited to two their types of fever. or three options, and only directive-style prompts were used in order to avoid command ambiguity”. They 2. A REVIEW OF RELATED WORK “partnered with Development Support Center (DSC), an NGO in Ahmedabad, Gujarat, to conduct a joint needs- Aditi et al. designed a speech dialogue system for the finding exercise, based on which the three system’s provision of health information to caregivers of HIV features were identified and implemented. Both isolated positive children in Botswana, Southern Africa. They word speech and DTMF versions of Avaaj Otalo were compared touchtone (DTMF) and speech input implemented. Prompts were recorded in a professional modalities in the context of low literacy users and a studio by one of the DSC radio program’s popular female health information service [3]. They hypothesized the voice personalities. Barge-in input was disallowed for following: both speech and DTMF. a. DTMF is likely to be more acceptable in the developing world because general numeracy is Avaaj Otalo was built and deployed using IBM Research less common; and India’s WWTW platform. Gujarati commands were b. (b) DTMF-based systems are much easier to converted to lexicons using the American English develop than natural language systems and are phoneme set for the speech recognition. The system consequently more attainable in the resource- performed with a recognition accuracy of 94% and was constrained environments that typically tested with 45 participants recruited from ten districts characterize the developing world. throughout rural Gujarat and they were all farmers. 87% of the subjects reported to have never used a PC”. The experiment was, however, not a within-subjects experiment design. “Input modality (speech vs. DTMF) was randomly assigned to each user, but was anonymously corrected to maintain balance across age, education and gender. Testing sessions were led by a DSC staff member who had experience communicating with the target user group.

In developing the system, doctors, nurses and Each participant completed three tasks with Avaaj Otalo nutritionists were interviewed as to the contents provided corresponding to its three features (listening to by the system. Further requirements were also got from announcements, listening to archived radio program

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

recordings, and posting questions), ordered by increasing participants used a regular wired office telephone on the difficulty. 38 participants were tested in a quiet office table. The speech output from the system was played using a landline phone, with only the DSC staffer and through a set of speakers as well as the handset so that two researchers as observers while the remaining 7 were the system’s output could be captured via a video camera women and were tested in their homes using a mobile in the usability lab. In the study, all three elements of phone because of travelling difficulty. The prototype usability of telephony applications according to the application was instrumented to log task completion, European Telecommunications Standards Institute’s errors and call duration to measure performance. During guidelines on usability in telephony applications were the test, two researchers noted points of difficulty, facial examined. The system effectiveness was measured by expressions, and comments made during the call. A post- calculating a success rate for each user task. The amount test questionnaire with Likert scales was administered to of time to finish each task was used as a proxy measure measure user satisfaction, ease of use and for system efficiency. An evaluation of user satisfaction learnability”[4]. was carried out through a series of questions asked immediately following the use of each modality” [5]. The results showed that “task completion rate with Their results indicated that “(a) DTMF was more DTMF was significantly higher than with speech (74% effective and efficient for linear tasks, whereas speech vs. 61%; p< 0.05). There was no significant difference in was better for nonlinear tasks; (b) speech was preferred to user satisfaction. In both groups, over 80% of users DTMF by a majority of users; (c) speech was judged as reported that they found it easy to access information being more satisfying, more entertaining and easier to use from the system. Over 75% of both groups said they than DTMF; and (d) user preference for a particular would “definitely” use such an application if it was made modality was better predicted by user performance in available. The users were asked to state the difficulty nonlinear tasks rather than linear ones” [5]. level of a particular task as either “difficult” or “very difficult” on a five-point Likert scale. Across all tasks, 3. METHODOLOGY the percentage of such responses was 49% for speech and 30% for DTMF (p < 0.05). For difficulty faced when HDES was developed using VoiceObjects Desktop for providing input to the system, 81% of DTMF users Eclipse 9 and Voxeo Prophecy 8 was used as the answered “no” or “definitely no”, compared to 38% for implementation platform. XLite softphone was used to speech users (p< 0.01)” [4]. call HDES for testing. All these tools allow for developing and testing dialogue systems on a PC without Kwan and Jennifer also reported “an experiment that having to deploy them on any telecoms service provider’s critically tested user’s preference for an input modality network. HDES could not be hosted as there is currently (speech vs. DTMF) in a phone-based message retrieval no single voice service provider that provides such system using a fully functioning natural language system. service in Nigeria. The experiment was a within-subjects design. All participants used both the speech-only and the DTMF- 3.1. Dialogue Flow of HDES only modalities and they all completed identical tasks for Once HDES is called, the system responds by welcoming both input modalities. Sixteen participants (8 women and the caller and introduces the caller to the services it 8 men) were recruited from the IBM T. J.Watson provides. The system then lists all the symptoms the Research Center located in the state of New York. The caller can choose from, as well as their corresponding participants were in a wide range of ages (from early 20s keys (Table 1). The caller has the option of making the to over 60). None of the participants had any prior system repeat all the symptoms as many times as possible experience using mobile assistant, and they were all until he is familiar with them. After this, the caller can native speakers of English. The participants took the say three initial symptoms or press the corresponding study one at a time in a usability lab. Upon arrival at the keys on the telephone keypad of the symptoms he has lab, the participants were seated and given a booklet with noticed he has. The caller would then need to confirm the an instruction on the first page. The instruction page symptoms supplied before diagnosis is carried out. If all described the purpose of the study and had a list of the symptoms are correct, the system goes on to required tasks. It also had a phone number they should diagnose, otherwise, the caller can re-supply or correct call to reach the system, the name of the test account they any of the initial three symptoms supplied. If the should use, and a password. In the case of the DTMF symptoms supplied are sufficient to diagnose the caller’s condition, the instruction page had the function mapping disease, the system responds by telling the caller the and a sample interaction. disease being suffered from. Otherwise, the system would need to ask a series of questions in order for it to determine the kind of ailment the caller is suffering from.

In the case of the speech-only condition, this page had sample phrases that could be spoken to the system. The Table 1: Symptoms and Their Equivalent DTMF Keys

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Key Symptom 1 Vomiting 2 Headache 3 Fever 4 Chills 5 High fever 6 Muscle pain 7 Tiredness/Fatigue 8 Sore throat 9 Diarrhea 10 Loss of appetite 11 Stomach pain 12 Rash 13 Bloody stool 14 Nose bleed 15 Low consciousness 16 Low heartbeat 17 Constipation 18 Bleeding from anus 19 Mouth bleeding 20 Joint pain 21 General discomfort 22 Abdominal rash 23 Chest rash 24 tongue discoloration 25 Excessive thirst 26 Black stool 27 Internal heat 28 Back pain 29 Dry cough 30 Cough 31 Tiredness 32 Weight loss 33 Chest pain 34 Blood pressure changes 35 Hypertension 36 Swollen neck 37 Swollen face 38 Swollen eyes 39 Ringing in ears 40 Blood in urine 41 Change in heartbeat 42 Shaking 43 Nausea 44 Dizziness 45 General pain 46 Sweating 47 Fall in temperature 48 Paleness 49 Shortness of breath 50 Bitter taste 51 Low urine 52 Stress 53 Anxiety 54 Skin redness 55 Red eyes 56 High heartbeat 57 Low backache 58 Red tongue 59 Bad breath odour 60 Bone pain 61 Blood vomiting

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

3.2. Experimental Design The subjects tested the system one at a time in an office. A paper list containing the symptoms as well as their The tools used for the implementation allow the testing of corresponding DTMF keys was given to each participant. the system in a standalone mode. They were taught how This was necessary because of memory loss associated to initiate a call to HDES using XLite soft phone. The with dialogue systems with many menu options and subjects were informed that the purpose of the study was correlates with Kwan and Jennifer’s approach. Barge-in to examine usability issues related to the use of a phone- was therefore allowed once a caller knew what key or based health dialogue expert system for diagnosing symptom to supply, so that there would not be any need diseases. Once each subject arrived at the venue where to wait till all the symptoms have been listed before the test was carried out, they were given a detailed responding. A within-subjects design was adopted in the instruction about what the system does and how it could evaluation as in the case of Kwan and Jennifer. All the be interacted with. A paper list containing the symptoms subjects tested both DTMF and speech separately. They and their equivalent DTMF keys was also given to them. completed the same task for each of the input modalities. The subjects were told that their task was to interact with Once a subject completed a task using a particular input HDES using the soft phone in order to diagnose their modality, he/she would fill the questionnaire for that ailments. They were asked to pick specific symptoms input modality. Once that was done, he/she would use the representing the ones they assumed to have and say or other input modality, after which the questionnaire on press them on the telephone keypad. They were informed that input modality would be filled. that they would need to fill a questionnaire each for each input modality. The speech output from the system was 3.3. Participants played through a set of external speakers connected to an Twenty one subjects participated in the evaluation. This Acer laptop running Windows Vista on which HDES number is more than 16 used by Kwan and Jennifer. The resided. participants consisted of undergraduates of Covenant University, Ota and University of Agriculture, Abeokuta, 3.6. Measures all in Nigeria. Others are doctors from Victory Medical As reported by Kwan and Jennifer, according to ETSI, Centre, Eleyele, Ibadan and State Hospital, Jericho, usability in telephony applications is defined as the level Ibadan, all in Oyo State, Nigeria. The doctors were five of effectiveness, efficiency, and satisfaction with which a in number. The subjects also included professionals in specific user achieves goals in a particular environment. teaching and health maintenance organizations. Nine of In this work, the three dimensions of usability were the subjects are males while ten are females. Two did not examined as in the case of Kwan and Jennifer. System indicate their gender. Five of the subjects are in the age effectiveness was measured by calculating the error rate range of 31- 40, four in the range of 21 – 30 and ten in for each of the input modalities. This was done by the range of 15 – 20. Two did not specify their age range. programming HDES to log errors: Misrecognition and No Match/No Input. Call duration was used as a 3.4. Apparatus measure of efficiency. This was extracted from a system 3.4.1. DTMF Modality log file after each user finished with an input modality. The subjects were instructed about the functionality User satisfaction was evaluated through the items in the provided by the system and how they could interact with questionnaires administered after the use of each input it. They were informed that their mode of interaction with modality. the system in this modality was through the pressing of the keys on the telephone keypad. The keys representing In measuring user satisfaction, a questionnaire used in a the symptoms were to be pressed, and all other similar study by Kwan and Jennifer [5] and user interactions with the system should be through the keys satisfaction survey used in Marilyn et al. [11] were instead of speech. For other interaction with HDES, they adopted. “The measures used in the questionnaire have were asked to listen to the system for guide at each stage both face and content validities. In terms of face validity, of the dialogue. Like the speech modality condition, the all measures were constructed by experts who have more system output was synthesized lady’s speech. than 10 years of experience in usability tests of mobile and speech user interface (SUI) applications. In terms of 3.4.2. Speech Modality content validity, the measures cover all dimensions of In the speech modality, the subjects were instructed to usability in telephony applications as defined by ETSI” use only speech for interacting with HDES. The list [5]. The only modification done to the questionnaire by containing the symptoms served as a guide for them as Kwan and Jennifer was the changing of certain adjectives regards the symptoms. They were, however, also asked to to their simpler synonyms so as to aid the subjects’ listen to the system for guide for other conversations with understanding. Two types of questions were administered it. – questions on modality evaluation and questions on system evaluation as adapted from Kwan and Jennifer [5]. “The first set of questions asked the subjects to evaluate 3.5. Procedure the system regardless of their evaluation of the

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

interaction modality”. These were extracted from Marilyn Satisfaction ratings for DTMF were significantly higher et al. [12, 13]. For the second set of questions, the (M=37.1) than for speech (M=33) as indicated by a subjects were to respond to questions that evaluated their significant t-test, t(18)= -3.45, tcrit =2.1, p < .05. Since - interaction modality with the system. It was speculated 3.45 is less than -2.1, the null hypothesis is rejected. The as done by Kwan and Jennifer that the evaluation of the finding thus indicates that DTMF was more satisfying interaction modality and the evaluation of the system than speech. could be different. 4.1.2. Modality Evaluation After carrying out the task, the subjects were immediately For modality satisfaction, the null hypothesis is that the “asked to evaluate the system and the interaction mean difference between DTMF and speech modalities modality by indicating how well certain adjectives satisfaction is zero. The alternative hypothesis is that described the modality and the system on Likert scales of there is a mean difference between the two input 1 (strongly disagree) to 5 (strongly agree)”. The modalities. adjectives – comfortable, uncomfortable (instead of exhausting used by Kwan and Jennifer ), frustrating and H0: µd =0 satisfying were used to measure modality satisfaction H1: µd ≠0 index. Four adjectives – boring, cool, fun, and entertaining – were used for a modality entertainment The subjects evaluated their interaction with speech as index. The adjectives used for measuring naturalness more satisfying (M=12.42) than DTMF (M=12.21) as index were- natural, unnatural (instead of artificial indicated by a t-test, t(18)= -.43, tcrit =2.1, p < .05, though used by Kwan and Jennifer), boring (instead of the two systems did not differ significantly. Since -.43 in repetitive used by Kwan and Jennifer and nervous not less than -2.1, the alternative hypothesis is rejected. (instead of strained used by Kwan and Jennifer [5]. Thus, the two are modalities are equally satisfying. Lastly, the subjects were required to provide answers to For modality entertainment, the null hypothesis is that the the question Which of speech and DTMF do you prefer mean difference between DTMF and speech modalities to interact with the system and why? entertainment is zero. The alternative hypothesis is that there is a mean difference between the two input 4. RESULTS modalities. As earlier mentioned, system effectiveness was measured by calculating the error rate for the task carried out using H0: µd =0 each of the input modalities. This was done by H1: µd ≠0 instrumenting HDES to log errors: Misrecognition and No Match/No Input. Call duration was used as a An interesting result was got for modality entertainment; measure of efficiency. User satisfaction was measured both speech and DTMF were rated equal (M=13.32) as through questionnaires. Subjects’ DTMF satisfaction indicated by a t-test, t(18)=0, tcrit =2.1, p < .05. Since 0 is ratings were compared with their speech satisfaction not less than -2.1, reject the alternative hypothesis. ratings using a repeated measures t-test with modality as Hence, the result shows that both speech and DTMF are the repeated factor. “The within-subjects t-test, used for equally entertaining. Finally, for modality naturalness, comparisons with a continuous dependent variable, is the null hypothesis is that the mean difference between also known as the paired samples t-test (the SPSS term), DTMF and speech modalities naturalness is zero. The the dependent samples t-test, correlated samples t-test, or alternative hypothesis is that there is a mean difference the repeated measures t-test. It is used when the same between the two input modalities. person is in the study twice”[14]. t-test is used when the sample size is less than 30[15]. H0: µd =0 H1: µd ≠0 4.1. User Satisfaction 4.1.1. System Evaluation The subjects evaluated their interaction with speech as The null hypothesis is that the mean difference between more natural (M=l0.63) than DTMF (M=10.12) as DTMF and speech user satisfaction is zero. The indicated by a t-test, t(18)=-.66, tcrit =2.1, p < .05. The alternative hypothesis is that there is a mean difference difference is not, however marginally significant. Since - between the two input modalities. .66 is not less than -2.1, the alternative hypothesis is rejected. Thus, DTMF and speech are equally natural. H0: µd =0 H1: µd ≠0

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4.1.3. Modality Preference In the use of devices for enhancing work, 5% replied that In response to the question Which of speech and DTMF they do not use technology to enhance their work. 79% do you prefer to interact with the system and why? 62 use laptop/notebook to enhance their work, while 16% % of the subjects chose DTMF modality, whereas 38 % use personal digital assistant (PDA)/cell phone. All the chose speech (p < .05). 67% of those who chose DTMF subjects reported that they owned a mobile phone or a gave recognition inaccuracy of speech modality as being PDA. 63% have owned a mobile phone/a PDA for more the reason why they preferred DTMF. Other gave reasons than 2 years, 16% for 6 months, 16% for 2 years and 5% like convenience of using DTMF, higher speed of task for 1 year. In response to the number of times they make completion with DTMF and that it is easier. For those or receive calls a week, 5% do not make and receive that chose speech, the reasons given are “It is inviting, calls. 11% make or receive calls 1-2 times a week, 16% interesting and exciting though not accurate”, “It is 3-4 times a week, 5% 5-6 times a week and 63% more easy to understand”, “It reduces stress”, “For clarity than 7 times a week. Lastly, 84% support the use of purpose”, “It is more natural”, “It is easier to use” and mobile devices for healthcare, while 16% do not. The “It is natural though not accurate”. 16% that do not support gave reasons such as speech misrecognition problem and that the use of mobile 4.1.4. Effectiveness and Efficiency devices will not be as fast as with an encounter with a For effectiveness and efficiency, only data from subjects medical practitioner. who were able to successfully interact with HDES using the two input modes were used. For modality 5. CONCLUSION effectiveness, the null hypothesis is that the mean difference between DTMF and speech modalities A within-subjects comparison between DTMF and effectiveness is zero. The alternative hypothesis is that speech for interacting with dialogue expert system for there is a mean difference between the two input diagnosing diseases among Nigerian users has been modalities. presented. The results presented in section 4 have implications for developers and it is clear that if H0: µd =0 effectiveness and efficiency are the focus, DTMF is H1: µd ≠0 recommended. If satisfaction is the watchword for the overall system, DTMF is recommended. Speech is DTMF was more effective than speech as the errors recommended for modality satisfaction while both generated by speech were significantly higher (M=8) than modalities are recommended for entertainment purpose. those generated by DTMF (M=2.5) as indicated by a Speech is advocated for modality naturalness, however, a significant t-test, t(7)=-2.26, tcrit =2.36, p<.05. The null well-designed diseases diagnosis dialogue system should hypothesis is therefore rejected. strategically provide a platform that incorporates the two input modalities to reap the benefits of the two, and to For modality efficiency, the null hypothesis is that the allow the users varieties of choices that best suit their mean difference between DTMF and speech modalities needs. efficiencies is zero. The alternative hypothesis is that there is a mean difference between the two input modalities.

H0: µd =0 H1: µd ≠0

DTMF was also more efficient (M=1.14) than speech (M=1.97) as the completion time is less than that of speech as indicated by a significant t-test, t(9)=-4.29, tcrit =2.26, p<.05. The null hypothesis is therefore rejected.

4.1.5. Experiences with Mobile and Computing Devices 5% of the subjects rated themselves as novice in the use of computer software. 16% rated themselves as expert. 68% rated their skill as being good, while 11% rated their skill as being average.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFERENCES [11] M. Walker, D. Litman and C. Kamm, “Evaluating [1]. R. José, “Web services and speech-based spoken language systems,” Proc. American Voice applications around VoiceXM,” Journal of Input/Output Society (AVIOS). May 1999. Networks ,vol. 2, February 2007, pp. 27-35. [12]. W. Marilyn et al. “PARADISE: A framework for [2]. W3C (2001). Voice Extensible Markup Language evaluating spoken dialogue agents,” Proc. 35th (VoiceXML) Version 2.0 Annual Meeting of the Association of http://www.w3.org/TR/2004/REC-voicexml20- Computational Linguistics (ACL 97), July 1997, 20040316/. Accessed May 25, 2008. pp. 271-280. [3]. S. G. Aditi et al., “HIV Health Information Access [13]. M. A. Walker, C. K. Kamm and D. Litman,. Using Spoken Dialogue Systems: Touchtone vs “Towards Developing General Models of Usability Speech,”. Proc. 3rd International Conference on With PARADISE,” Natural Language Engineering, Information and Communication Technologies and vol. 1, 1998, pp 1-13. Development, Doha, Qatar, 2009, pp. 95 – 107. [14]. Newsom USP 534 Data Analysis Spring 2009 [4]. P. Neil et al., (2009). “A Comparative Study of http://www.upa.pdx.edu/IOA/newsom/ho_t- Speech and Dialed Input Voice Interfaces in Rural test%20within.pdf. Accessed March 31, 2011. India,” Proc. CHI 2009, Boston, MA, USA, April 3 [15]. G. B. Alan, Elementary Statistics: A Step by Step - 9, 2009. Approach, New York: Mc GrawHill, 2004. [5]. M. L. Kwan and L. Jennifer,.“Speech Versus Touch: A Comparative Study of the Use of Speech and DTMF Keypad for Navigation,” International Journal of Human–Computer Interaction, vol. 19, 2005, pp. 343–360. [6]. O. M. Olufemi, U. C. Onuwa and A. A. Taofeek “Integration of Expert System Technology Into VoiceXML-Based Systems,” Journal of Computing, vol. 3, 2011, pp. 2151-9617. [7].World Health Organization (2010). World Health Statistics 2010. [8]. F. O. Ogunrin, O. Ogunrin and A. Akerele, “Motivating Nigerian Doctors for Improved Healthcare Delivery,” International Journal of Health Care Quality Assurance, vol. 20, 2007, pp. 290-306. [9]Nigerian Communication Commission (2012). Monthly Subscriber Data http://www.ncc.gov.ng/index.php?option=com_c ontent&view=article&id=125&Itemid=73. Accessed December 20, 2012. [10]. National Population Commission (2012). http://www.population.gov.ng/.Accessed December 20, 2012.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Development of an Automated Parking Lot Management System

Segun O. Olatinwo Department of Computer Science and Engineering Ladoke Akintola University of Technology Ogbomoso, Nigeria. Department of Computer Engineering , Moshood Abiola Polytechnic, Abeokuta, Nigeria. [email protected] and [email protected]

O. Shoewu Department of Electronic and Computer Engineering Lagos State University Epe Campus, Lagos State, Nigeria [email protected]

ABSTRACT This paper attempts to design and implement an automated parking lot management system. Automated Parking Lot Management System is a fully functional and digitally controlled parking lot management system that is implemented with the use and integration of different digital circuitry and micro computing. The design involves different stages, from the main unit, process is passed on to different subunits to achieve the goal of full automation. An oncoming car will communicate (through the driver) wirelessly with the main unit attached to the Parking Facility Gate. The main unit will verify the transmitted access information and will pass control after verification to the gate mechanism drivers, this in turn drives the right gate control (either exit or entry unit). The system now monitors the activity of the driver afterwards, and for entry, as the driver moves a predetermined distance into the facility ,the system turns back the gate mechanism (for closure of gate) and passes control to the space allocation and management unit.The objective of this later unit is to manage the parking spaces available in the lot by monitoring the activity of the cars inside, allocating the spaces in an orderly manner, monitoring compliance and notify the overall control center (manned) of the space(s) available. It has a display interface for communicating with the users of the facility. There is also a control center that is manned by personnel and monitors the activities within the parking lot. It is notified of any activity, space(s) available and also the overall system can be shut down or switched on from the control center.The main goal of this project is to achieve full automation and it will find immediate usage in large facilities with different access restrictions, government properties, and university campuses to sectionalize lecturer’s car park and student’s car park etc.

Keywords: Microcontroller, Microchip, Sensors, Ground Switch, Control Center, AT89C52

African Journal of Computing & ICT Reference Format Segun O. Olatinwo & O. Shoewu (2013). Development of an Automated Parking Lot Management System. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 95-108

1. INTRODUCTION

The purpose of this project is to create an automated Due to the fact that needs will vary, this paper attempts to parking lot management system that is easy to operate develop a model highly suited for non-commercial and also does not give away anything in terms of parking lots. performance. Automating a parking facility provides many advantages and can basically be used in public and 2. RELATED WORKS commercial premises. It is effective for controlling access to different areas of a large facility and has a particular Several research work and development have been advantage in that it can be programmed to suit different involved overtime; say for two decades now, in operational requirement and security details. The use of a developing various types and classes of automated gate fully automated system enhances and contributes to the operation systems which have met the diverse needs of prestige of the protected site when it comes to using individuals all over the world. However such systems technology to enhance property value. It can also be used operates only at the entry and exit points, they do not on university campuses to control parking areas. monitor the activities within the garage and manage the parking spaces.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Most previous only focuses on the entry and exit and the The gate unit will be implemented with the use of swing designs are hardwired (i.e. not easy to adopt for different gates and controls, which will be driven by a linear systems). This system being developed, apart from actuator or electric motor coupled with the necessary controlling the entry and exit into the facility, monitors mechanical parts. The system will detect any entry, exit, the spaces available in the parking lot, allocates space in or refusal to enter or exit (after activating gate controls) particular order, automatically detect an empty parking and also activate the closure of the gate arm when space, shuts down the system when all spaces are used necessary. The parking spaces will be allotted in and also can be controlled and monitored from a control ascending order and activity on each space will be center. The main features considered when constructing monitored. The system can also be fully SHUT DOWN this automated management system are basically the from a control center and turned ON at the push of a efficiency of the control system, and its reliability. Also button from the same center. There would also be control the strength of the system is another area which needs to indicators signaling the entry, exit, shut down, turn on be adequately catered for in other to improve the security. and the parking spaces remaining. The reliability of the automated entry and exit parking lot system will be confirmed on testing of the constructed 3. METHODOLOGY automated system model while the strength will be worked on in the real-life deployment. The efficiency of This paper attempts to construct an automated parking lot the control system on the other hand, is fully dependent management system that can be deployed for use in on the quality of the components used. diverse premises. The use of digital circuitry and micro computing is to be employed to achieve this fully With the controllers and other driving circuits beneath, automated system. The PIC 16F84A would be use to we hope to achieve a fully functional system that allows code signal to be transmitted with the appropriate the driver (car or allowed personnel into a parking transmitter and a receiver module with AT8952 facility), to communicate with the gate (via gate sensor Microcontroller will serve as the as the heart of the main and circuit) and pass on process control to other units. unit. The microcontrollers will be programmed using C Decision can either be to open the gate, to leave the gate Language. Varying TTL gates will be used for parking closed or to alert security personnel. On entering the space(s) management (allocation and monitoring). A facility, the parking space management system takes over brief flowchart is shown in Figure 2. and allocates empty space to the user. A pictorial representation of how the system works is given in figure 1.

Figure 1: Block diagram of the proposed system

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

START 4. SYSTEM DESIGN

START The design of the ‘automated Parking Lot management System’ will be in three stages. Stage one deal with the transmission of programmed signal from the car while NO stage two discusses the receiving of transmitted signal CLOSE KEY=0 ? and also signal processing and operation control, stage three however discusses the space management and allocation. There are numerous design options and some YES come with deeper compromises in terms of safety, efficiency and ease of use without leaving out durability. We attempt to examine the numerous design problems INPUT FROM A SENSOR with the different design options available, before we arrived at a plausible conclusion. We arrived at a fully functional system that works in the following sequence:

1. The oncoming car transmits certain codes (entry codes) to the gate module which NO VALID INPUT ? compares the transmitted signal with the preset requirement. 2. The gate module on receiving the signal, YES processes it and releases the driving mechanism INITIALISE GATE OPEN OR CLOSE (gate driving mechanism), or remains shut in FUNCTION case of invalid access codes detection. 3. As the car enters the facility, it triggers a ground switch which returns the gate mechanism for closure of the gate arm. START COUNTDOWN TIMER 4. On entering the facility, the space management unit allocates empty space to the incoming car. 5. There will be a system override control at a control center which can halt or switch ON the INPUT FROM GROUND whole system. It also receives a real time SWITCH update on the parking space(s) available from the space management unit. NO The above processes will work same way for any car exiting the parking lot. Therefore I will concentrate on the entry design and duplicate the design at the exit. TIMER=0 INPUT = 1 In the following section, we presents the design of the ? transmitting module.

YESd 4.1 The Transmitting Module INITIALISE GATE CLOSE FUNCTION. The design of this module is broken down into the following areas:  design problems  design techniques

 design analysis RETURN  design integration and simulation, and  Theory of output.

4.2 Design Problems Designing an effective transmitting module requires strict adherence to certain procedural and operational requirement in order to obtain a transmitter that is Figure 2: Flowchart of the proposed Parking Lot efficient, durable, and most importantly, easy to install. Management System The transmitted signal should also get to the receiving end without interference, since this can come from various sources. Also

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

1. The transmission must be fast, less than or Advantages of Infrared equal 1000ms. These include: 2. For security reasons, the transmitter can only 1. Low power requirements be activated by the driver. 2. Low circuitry costs for the entire 3. For efficiency, the range of transmission should coding/decoding circuitry be between 1-5 meters. 3. Simple circuitry: no special or proprietary hardware is required, can be incorporated into 4.3 Design Techniques the integrated circuit of any device In order to achieve and solve the necessary design 4. Higher security: directionality of the beam problems, the system was implemented as follows: helps ensure that data isn't leaked or spilled to 1. Transmission was done through Infrared nearby devices as it's transmitted medium to achieve a transmitted signal free of 5. Portability Electromagnetic interference from a car engine and other sources of Electromagnetic Radio Frequency Disadvantages: Interference. 1. Interference: communication devices using 2. Used microcontroller PIC16F84A to program similar frequencies - wireless phones, scanners, the transmitted signal because of its portability wrist radios and personal locator's can interfere and flexibility in terms of operation with transmission 3. 555 timer was used since the infra red 2. Lack of security: easier to "eavesdrop" on transmitter will be transmitting the modulated transmissions since signals are spread out in signal, (pulse width modulation), at a carrier space. frequency approximately equal to 40KHZ 3. Higher cost than infrared. (transmitting IR at this frequency help 4. Requires government licensing eliminate interference) 5. Lower speed: data rate transmission is lower 4. Power supply to this unit will be from the car than wired and infrared transmission battery stepped down to 5Vdc with the use of an analog IC or the device could be powered The Advantages of Radio Frequency over IR has to do with the use of battery cells. with: 1. No need for line of Sight 4.4 Design Analysis 2. Longer range The Infrared (IR) Advantage 3. No interference from other sources of light like It is an invisible band of radiation at the lower end of the sunlight. visible light spectrum. With wavelengths from 50 nm to 1 mm, infrared starts at the end of the microwave spectrum These disadvantages of IR was taken care of, such that and ends at the beginning of visible light. Infrared the radiating angle of the IR emitter over the short range transmission typically requires an unobstructed line of required (1-5m) is enough to avoid weak signal at the sight between transmitter and receiver. In infrared receiver. The interference with other sources light was communication an LED transmits the infrared signal as also effectively taken care of using the SIRC protocol bursts of non-visible light. At the receiving end a photo with the transmission of signal. diode or photo receptor detects and captures the light pulses, which are then processed to retrieve the 4.5 The Sirc Protocol information they contain. In a bid to eliminate ambient light sources, both natural and man made, from interfering with other data stream Widely used in most audio and video remote controls, transmitted by the handset, modulated light is used. This infrared transmission is also used for wireless modulation is centered on frequencies depending on the connections between computer devices and a variety of manufacturer; and varies from 32 kHz to 40 kHz. In the detectors and control machinations. Infrared technology case of some devices, the modulation is centered at 40 offers several important advantages as a form of wireless kHz, which means we require a device that can receive communication. Advantages and disadvantages of IR are the modulated infrared light and convert it into a TTL first presented, followed by a comparative listing of radio signal that the micro controller can handle. There are a frequency (RF) advantages and disadvantages. This number of these devices available, each having a specific comparison will show why I prefer to use the Infrared center frequency that they are more sensitive too. In a bid instead of Radio Frequency (RF). to eliminate ambient light sources, both natural and man made, from interfering with the data SIRC (Serial Infra- Red Control) uses a form of pulse width modulation (PWM) to build up a 12-bit serial interface, known as a packet. This is the most common protocol, but 15-bit and 20-bit versions are also available. A pulse with duration of 2.4ms is sent first as a header, this allows the internal

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

AGC to adjust and also allows the receiver to check if a • Only 35 single word instructions to be used valid packet is being received. A 1-bit is represented by • All instructions single cycle except for program pulse duration of 1.2ms, while a 0-bit has duration of branches which are two-cycle 0.6ms. A delay of 0.6ms is placed between every pulse. • Operating speed: DC - 20 MHz clock input DC - 200 ns instruction cycle • 1024 words of program memory • 68 bytes of data RAM • 64 bytes of data EEPROM • 14-bit wide instruction words • 8-bit wide data bytes • 15 special function hardware registers • Eight-level deep hardware stack • Direct, indirect and relative addressing modes Figure 3: The string of pulses for 12-bit packet • Four interrupt sources: - External RB0/INT pin - TMR0 timer overflow The string of pulses build up the 12-bit packet consisting - PORTB<7:4> interrupt on change of a 5-bit (0...31) device code, which represents a device, - Data EEPROM write complete and a 7-bit (0...127) button code, which represents the actual button pressed on the remote. The packet is Peripheral Features: transmitted most significant bit first (MSB), with the • 13 I/O pins with individual direction control device code being sent, then the button code. After the • High current sink/source for direct LED drive packet is sent, a delay is implemented, which brings the - 25 mA sink max. per pin whole transmitted signal to a length of 45ms. - 25 mA source max. per pin • TMR0: 8-bit timer/counter with 8-bit programmable prescaler

5. DEVICE OVERVIEW

With the right protocol aiding an interference free The PIC16F84A belongs to the mid-range family of the transmission, the IR is the best means of transmission for PICmicro™ microcontroller devices. A block diagram of the system. the device is shown below:

The program memory contains 1K words, which 4.6 Microchip Pic16f84 Microcontroller translates to 1024 instructions, since each 14-bit program The PIC16F84 is an 18bit- enhanced flash/EEPROM 8- memory word is the same width as each device bit micro controller with the following features: instruction data memory (RAM) contains 68 bytes. Data EEPROM is 64bytes.There are also 13 I/O pins that are user-configured on a pin-to-pin basis. Some pins are multiplexed with other device functions. These functions include: • External interrupt • Change on PORTB interrupt • Timer0 clock input

5.1 Manipulating The Microcontroller The microcontroller is basically a hardware device that requires programming to change its functionality. Manipulating the PIC require the use of a chosen software language. Choices can vary from low level Figure 4: PIC 16F84 languages to high level languages. The most common programming tool is the use of C language- an high level language. Using C affords one the opportunity to program the microcontroller from a perspective that reduces the codes and instructions to an easy to read and understand code. The programming was done using C language.

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

5.2 THE 555 TIMER 555 TIMER OPERATION MODES The 555 timer is one of the most remarkable integrated 2) Mode A: Astable operation circuits ever developed. It comes in a single or dual When configured as an oscillator the 555 timer is package and even low power CMOS versions exist - configured as in figure below. This is the free running ICM7555. Common part numbers are LM555, NE555, mode and the trigger is tied to the threshold pin. At LM556, and NE556. The 555 timer consists of two power-up, the capacitor is discharged, holding the trigger voltage comparators, a bi-stable flip flop, a discharge low. This triggers the timer, which establishes the transistor, and a resistor divider network capacitor charge path through Ra and Rb. When the capacitor reaches the threshold level of 2/3 Vcc, the output drops low and the discharge transistor turns on. The timing capacitor now discharges through Rb. When the capacitor voltage drops to 1/3 Vcc, the trigger comparator trips, automatically re triggering the timer, creating an oscillator whose frequency is determined by the formula in figure below.

Figure 5: 555 Timer

1) PIN FUNCTIONS Ground (Pin 1): Not surprising this pin is connected directly to ground.

Trigger (Pin 2): This pin is the input to the lower comparator and is used to set the latch, which in turn causes the output to go high.

Output (Pin 3): Output high is about 1.7V less than supply. Output high is capable of I source up to 200mA while output low is capable of I sink up to 200mA.

Reset (Pin 4): This is used to reset the latch and return the output to a low state. The reset is an overriding function. Figure 6: Astable Mode When not used connect to V+.

Control (Pin 5): Allows access to the 2/3V+ voltage While the 555 timer will operate up to about 1 MHz it is divider point when the 555 timer is used in voltage generally recommended it not be used beyond 500 KHz control mode. When not used connect to ground through owing to temperature stability considerations. So I will be a 0.01 µF capacitor. using it at 40 kHz. In the astable mode.

Threshold (Pin 6): This is an input to the upper 6. DESIGN STAGES, INTEGRATION AND comparator. SIMULATION

Discharge (Pin 7): This is the open collector to Q14 in The design of the transmitting module is broken down figure 4 below. into different stages, namely:

V+ (Pin 8): This connects to Vcc which is between 3 –  Stage A: the signal coding and C programming 16Vdc  Stage B: the carrier frequency generation  Stage C: integration of design stages and simulation.  Stage D: connection to IR LED and controlling transmission distance. 

100

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Stage A: Signal Coding 4. An output voltage of 5Vdc represents a high To achieve a good level of security of transmitted signal, level voltage and 0V represents, low level the transmitted signal is encoded with certain information voltage that the receiver module wait to receive and process. With the use of C language, we encoded the PIC16F84 microcontroller (using the serial infrared protocol Stage B: Carrier Frequency Generation And Simulation described earlier) to produce an output signal. The In order to achieve a transmitted signal free from ambient transmitted signal will be 0x92, for simplicity and the light interference and other forms ofinterference, we used burst will be transmitted in a total time of 45mS. 555 timer to modulate the signal and transmit a signal centered at 40Khz frequency. This gives the transmitted signal special characteristics that is peculiar and therefore makes it free from interference. Setting up the 555 timer in astable mode, I will be using the RESET pin as the output to the 555 timer. A high level voltage (+3 to +5Vdc) on the pin turns on the timer and a low level voltage (0- 1.5Vdc) turns off the timer.

Figure 9: Output Waveforms

The outputs of the coding stage (is the output of the PIC16F84) acts as the input into the 555 timer, which means pin RB1 of the 16F84 is connected in the RESET Figure 7: Signal Coding pin of the 555 timer. The physical layout is as shown above. The 555 timer is turned ON for 1.2mS when q A bit 1 is represented by 1.2mS signal and a bit 0 is pulse of that width enters through the reset pin, the timer represented by 0.6mS long signal. Each bit is represented is turned off for 0.6mS when the RESET pin defects the by a gap lapse of 0.6mS. Using an electronic simulation 0.6mS long low voltage gap. The timer is turned back software, I simulated the design (microcontroller and ON for 0.6mS when the high level 0.6mS pulse gets to program) and got my desired output waveform. The the RESET pin. This variation in input in turn varies the waveform is as follows: output of the 555 timer accordingly.

. Stage C: Integration Of Design Stages And Simulation The rapid advance in the computer technology has made electronic simulation of designs and integration of electronic circuits a much easy task. Each stage as 45ms highlighted above has been simulated and all has produced the required result. Integrating the different Figure 8: Output waveform of coding stage design stages produces a circuit diagram as shown below.

Characteristics Of The Output Waveform 1. It is 45mS long 2. It's low level voltage bits is the gap lapse of 0.6mS. 3. The high level voltage bits has two (2) modes; viz: the 0.6mS long mode representing a bit 0, the 1.2mS long mode representing a bit 1.

101

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 10: Schematic Diagram of Transmitting Module

STAGE D: CONNECTION TO IR LED AND  There must be a system to recognize the entry CONTROLLING THE TRANSMISSION DISTANCE of the car, and another to reverse the operation The output of the 555 timer turns ON and OFF the IR of the gate mechanism. LED for the specific period of each pulse applied to it.  There must be control keys to either shut down This period (of each bit) is what is now transmitted by the or switch on the system. infra red emitter. When it is ON, it is either 0.6mS or  There must be a suitable time lapse for the gate 1.2mS and when it is off, it is for 0.6mS. The distance of mechanism control to fully turn in the required transmission is varied with the use of a resistor in parallel direction (for opening or closing). Also there with the IR Emitter. The resistance value can vary must be time lapse for the gate mechanism to depending on the Emitter used. The physical layout of the remain OPEN in order to allow for the passage integrated circuit is shown above in figure 9. of cars, (or turning back of car).

6.1 The Receiver Module  The system must have a fast response time not Design Overview exceeding 3000ms. This includes the time it In this section, we attempt to design a receiver module will take the arm to fully open. that will not only receive the transmitted signal, but also decode the signal and control the operation of the gate 6.3 Design Technique mechanism depending on the truth or otherwise of In order to achieve an entry system that will receive the transmitted signal after verification. Also it will pass on transmitted signal at 38-40KHz with ease, we employed a process to the space management unit. specially designed infrared sensor, that can only sense The general requirements of the entry system are as signals at that frequency range. I will also be using the follows: AT89S52 for decoding the received signal and controlling the overall operation of the gate mechanism.  The receiver sensor must operate or be The AT89S52 will work for both the entry and exit designed to recognize signal at 38-40KHz. systems. Therefore, I will be using a single  There must be a decoding circuitry to decode microcontroller for the gate module. In addition to this, in the transmitted signal and recognize the pulse order to sense the entry of a car ( or its exit), we made use burst. of a ground switch which will monitor the movement of  the car, while a special timing circuit in the  microcontroller will determine if a car is entering or

102

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

exiting after a period of time ( this is to prevent the gate By combining a versatile 8-bit CPU with Flash on a arm from remaining opened when driver changes his or monolithic chip the Atmel AT89C52 is a powerful her mind to enter or leave the facility).The microcomputer which provides a highly flexible and cost microcontroller will activate or deactivate a pair of relay, effective solution to many embedded control which will in turn supply or cut off voltage to an electric applications. The AT89C52 provides the following motor or linear actuator which will drive the gate arm. standard features: 8Kbytes of Flash, 256 bytes of RAM, The linear actuator (or electric motor) could be used to 32 I/O lines, three 16-bit timer/counters, a six-vector two- drive the mechanism controlling the arm (or sliding gate). level interrupt architecture, a full duplex serial port, on- A control center manned by the system administrator will chip oscillator, and clock circuitry which makes it ideal comprise of system turn ON and turn OFF switches and for the project. In addition, the AT89C52 is designed also light indicators signaling the entry or exit of a car with static logic for operation down to zero frequency into the facility. and supports two software selectable power saving modes. The Idle Mode stops the CPU while allowing the Power supply will be from the mains and would be RAM, timer/counters, serial port, and interrupt system to stepped-down as required for different stages of the continue functioning. The Power Down Mode saves the circuit. Side supply of power for the microcontroller and RAM contents but freezes the oscillator, disabling all higher voltage needed to drive the electric motor or linear other chip functions until the next hardware reset. actuator depending on the particular application will also be provided. Pin Description

7. DESIGN ANALYSIS (COMPONENT ANALYSIS) VCC: Supply voltage

WHY AT89S52? GND: Ground The AT89S52 is a low-power, high-performance CMOS 8-bit microcomputer with 8 Kbyte of flash programmable Port 0: Port 0 is an 8-bit open drain bidirectional I/O port. and erasable read only memory (PEROM). The device is As an output port, each pin can sink eight TTL inputs. manufactured using Atmel’s high density nonvolatile When 1s are written to port 0 pins, the pins can be used memory technology and is compatible with the industry as high-impedance inputs. Port 0 can also be configured standard 80C51 and 80C52 instruction set and pin out. to be the multiplexed low-order address/data bus during The on-chip Flash allows the program memory to be accesses to external program and data memory. In this reprogrammed in-system or by a conventional mode, P0 has internal pull ups. Port 0 also receives the nonvolatile memory programmer. code bytes during Flash programming and outputs the code bytes during program verification. External pull ups are required during program verification.

Port 1: Port 1 is an 8-bit bidirectional I/O port with internal pull ups. The Port 1 output buffers can sink/source four TTL inputs. When 1s are written to Port 1 pins, they are pulled high by the internal pull ups and can be used as inputs. As inputs, Port 1 pins that are externally being pulled low will source current (IIL) because of the internal pull ups. In addition, P1.0 and P1.1 can be configured to be the timer/counter 2 external count input (P1.0/T2) and the timer/counter 2 trigger input (P1.1/T2EX), respectively

PORT ALTERNATE PIN FUNCTIONS P1.0 T2 (external count input toTimer/Counter 2), clock-out P 1.1 T2EX (Timer/Counter 2 capture/reload trigger and direction control) Figure 11: AT89C52

103

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Port 1 also receives the low-order address bytes during no effect if the microcrontroller is in external execution Flash programming and program verification. Port 2: Port mode. PSEN Program Store Enable is the read strobe to 2 is an 8-bit bidirectional I/O port with internal pullups. external program memory. When the AT89C52 is The Port 2 output buffers can sink/source four TTL executing code from external program memory, PSEN is inputs. When 1s are written to Port 2 pins, they are pulled activated twice each machine cycle, except that two high by the internal pullups and can be used as inputs. As PSEN activations are skipped during each access to inputs, Port 2 pins that are externally being pulled low external data memory. will source current (IIL) because of the internal pullups. Port 2 emits the high-order address byte during fetches EA/VPP: External Access Enable: EA must be strapped from external program memory and during accesses to to GND in order to enable the device to fetch code from external data memory that use 16-bit addresses (MOVX external program memory locations starting at 0000H up @ DPTR). In this application, Port 2 uses strong internal to FFFFH. Note, however, that if lock bit 1 is pullups when emitting 1s. During accesses to external programmed, EA will be internally latched on reset. EA data memory that use 8-bit addresses (MOVX @ RI), should be strapped to VCC for internal program Port 2 emits the contents of the P2 Special Function executions. This pin also receives the 12-volt Register. Port 2 also receives the high-order address bits programming enable voltage (VPP) during Flash and some control signals during Flash programming and programming when 12-volt programming is selected. verification. XTAL1: Input to the inverting oscillator amplifier and Port 3: Port 3 is an 8-bit bidirectional I/O port with input to the internal clock operating circuit. internal pullups. The Port 3 output buffers can sink/source four TTL inputs. When 1s are written to Port XTAL2: Output from the inverting oscillator amplifier. 3 pins, they are pulled high by the internal pullups and Timer 2: Timer 2 is a 16-bit Timer/Counter that can can be used as inputs. As inputs, Port 3 pins that are operate as either a timer or an event counter. The type of externally being pulled low will source current (IIL) operation is selected by bit C/T2 in the SFR T2CON because of the pullups. Port 3 also serves the functions of (shown in Table 2). Timer 2 has three operating modes: various special features of the AT89S52. capture, auto-reload (up or down counting), and baud rate generator. The modes are selected by bits in T2CON. PORT PIN ALTERNATE Timer 2 consists of two 8-bit registers, TH2 and TL2. In FUNCTIONS the Timer function, the TL2 register is incremented every P3.0 RXD (serial input port) machine cycle. Since a machine cycle consists of 12 oscillator periods, the count rate is 1/12 of the oscillator P3.1 TXD (serial output port) frequency. In the Counter function, the register is P3.2 INT0 (external interrupt 0) incremented in response to a l-to-0 transition at its P3.3 INT1 (external interrupt 1) corresponding external input pin, T2. In this function, the P3.4 T0 (timer 0 external input) external input is sampled during S5P2 of every machine P3.5 T1 (timer 1 external input) cycle. When the samples show a high in one cycle and a P3.6 WR (external data memory write low in the next cycle, the count is incremented. The new strobe count value appears in the register during S3P1 of the cycle following the one in which the transition was P3.7 0 detected. Since two machine cycles (24 oscillator periods) are required to recognize a 1-to-0 transition, the Port 3 also receives some control signals for Flash maximum count rate is 1/24 of the oscillator frequency. programming and programming verification. To ensure that a given level is sampled at least once before it changes, the level should be held for at least one RST: Reset input. A high on this pin for two machine full machine cycle. cycles while the oscillator is running resets the device. ALE/PROG: Address Latch Enable is an output pulse for latching the low byte of the address during accesses to external memory. This pin is also the program pulse input (PROG) during Flash programming. In normal operation, ALE is emitted at a constant rate of 1/6 the oscillator frequency and may be used for external timing or clocking purposes. Note, however, that one ALE pulse is skipped during each access to external data memory. If desired, ALE operation can be disabled by setting bit 0 of SFR location 8EH. With the bit set, ALE is active only during a MOVX or MOVC instruction. Otherwise, the pin is weakly pulled high. Setting the ALE-disable bit has

104

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Auto-Reload (Up or Down Counter) Figure below shows the internal operation and block diagram of one of this device. RCLK + CP/RL2 TR2 MODE

TCLK 0 0 1 16-Bit Auto-Reload 0 1 1 16-Bit Capture 1 X 1 Baud Rate Generator X X 0 (off)

Timer 2 can be programmed to count up or down when configured in its 16-bit auto-reload mode. This feature is invoked by the DCEN (Down Counter Enable) bit located in the SFR T2MOD . Upon reset, the DCEN bit is set to 0 so that timer 2 will default to count up. When DCEN is set, Timer 2 can count up or down, depending on the value of the T2EX pin. In this mode, two options Figure 13: Operation and block diagram of ISI U60 are selected by bit EXEN2 in T2CON. If EXEN2 = 0, Timer 2 counts up to 0FFFFH and then sets the TF2 bit As you can see, this deceptively simple looking device is upon overflow. The overflow also causes the timer a lot more than a re-packaged IR Photo diode. It filters, registers to be reloaded with the 16-bit value in RCAP2H amplify and demodulate the infrared signal. Then give a and RCAP2L. The values in RCAP2H and RCAP2L are nice clean TTL output by means of a final comparator preset by software. If EXEN2 = 1, a 16-bit reload can be stage. It also have a built in automatic gain control triggered either by an overflow or by a l-to-0 transition at (AGC), which helps stops overloading; if the transmitter external input T2EX. This transition also sets the EXF2 is too close. Most IR sensors have an active low output, bit. Both the TF2 and EXF2 bits can generate an interrupt which means that the micro controller is presented with a if enabled. logic 0 when an infrared signal is detected. With no signal present, a maximum current of 4.8mA is consumed 7.1 The Infrared Sensor (2.8mA being typical). In a bid to eliminate ambient light sources, both natural and man made, from interfering with the data stream 7.2 The Use Of Relays transmitted by the car module, modulated light is used. A relay is an electromagnetic switch. In other words it is This modulation is centred around particular frequency activated when a current is applied to it. Normally a relay and varies from 32kHz to 40kHz. In the case of our is used in a circuit as a type of switch . There are specific application, the modulation is centred at 40kHz, different types of relays and they operate at different which means we require a device that can receive the voltages. For this project the type of relay use can vary modulated infrared light and convert it into a TTL signal based on the particular application. The main part of a that the micro controller can handle. There are a number relay is the coil at the centre. A small current flowing of these devices available, each having a specific centre through the coil in the relay creates a magnetic field that frequency that they’re more sensitive too. The device pulls one switch contact against or away from another. used for this project is the IS1U60 from Sharp. It has a Putting it simply, when current is applied to the contacts centre frequency of 38Khz, which is close enough to at one side of the relay the coil allows the contacts at the 40kHz so as not to matter. other side to work. Usually relays are used to turn on a second circuit. The first circuit activates the relay which then ‘turns on’ the second circuit. I will be using two relays for the entrance and another two for the exit.

Figure 12: ISI U60

105

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

8. STAGES DESIGN, INTEGRATION AND The ISI U60 on detecting a signal, sends a high or low SIMULATION pulse of 3-5Vdc or 0-1.5Vdc to the microprocessor for The design stages and the different outputs are discussed the period of the signal received and this is processed by in this section. They include: the microcontroller as a combination of 1's and 0's (bit 1  signal detection or bit 0).  signal processing 8.2 Signal Processing And Control Stage On receiving the transmitted signal from the ISO U60, 8.1 Signal Detection the AT89S52 decodes the signal after reconstructing the The detection of transmitted signal at 38-40KHz is received signal. It compares it with the control code achieved with the use of ISI U60. There is a unit for the stored in the microcontroller. The microcontroller sends entrance and another for the exit. The detected signal is signals to either turn ON the relay or alternately reverse sent to AT89S52 micro controller for decoding and the operation. processing.

Figure 14: Gate unit Schematic

The operation of the AT89S52 can be overidden only by Pin5 (P1.4) and Pin6 (P1.5) are connected to the ground an interrupt from the control pins switch on the switch that monitors the movement of the car. Activation microcontroller ( which switches off the system), or of these reverses the operation of the relays. The circuit switching ON the microcontroller (which switches ON physical arrangement is shown below: the system). Pin3 (P1.2) and Pin4 (P1.3) of the microcontroller gives the code for shut down of the system, activation, and entry and exit indicators respectively.

106

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

8.3 The Space Allocation Unit each parking space. There will be a feed signal to the The main purpose of the space allocating unit is to microcontroller close pin for the closure of the system manage the parking spaces available in the parking lot. It when lots are full. This unit will also notify the control allocates in an orderly manner, displays the allocated center of the spaces available so that appropriate action space through an LED display and monitors activities on can be taken. The schematic diagram for this unit is as shown on the below:

Figure 15: Schematic diagram for the space allocation unit

9. RESULT The system was tested in order to evaluate its For the receiver and the control unit, the following performance in relation to the expected performance. outputs are expected  Ability to receive infra red signal 9.1 Objectives and Theoretical Results transmitted between 38 – 40KHz For the transmitter unit, the expected outputs are as  Process the received signal ,for follows: validation  Signals coded using pulse width adjustments  Control the appropriate gate driver depending on the IR receiver activated  An infra red signal at centered between 38- 40KHz  Monitor the movement of car through a switch so that gate driver can be

reversed

107

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

 Ability to shut down the system or 11. RECOMMENDATION FOR PRACTICE power it as required from the control center(manned) The following suggestions should be considered in carrying out further work on this study:  Ability to reset the overall system The transmitter – in other to increase operation convenience and add greater value and security,  Feedback indicators to the control we would recommend the use of the car center. headlamp for the transmission of coded signal. This will be controlled/triggered by a hidden For the space allocation and management unit, the switch in the car or a switch as the key holder. following outputs are expected  Monitor the available spaces real time The electric motor (used to drive the gate) - we recommend the use of linear actuator  Allocates spaces in a defined order to and/or speed control circuits in other to achieve the incoming car controlled speed for the gate and also give it  Display the space allotted using seven resistance against manual push. segment displays Space allocation unit feedback- we  Notify the control center of space(s) recommend the feeding back of the output of remaining this stage to the microcontroller so that the

system can be automatically shut down when The above represents the theoretical results indicating the outputs of different stages of this work. facility is filled.

9.2 Actual Result After Construction/Comparison REFERENCES With The Expected Result [1] http:// www.samect.com, Introduction to car The actual results after the construction was compared security using PIC 16 F 84A with the expected outputs and we found out that the [2] Boylestard, R. and Nasheisky, Electronic results tally. device and Circuitry Theory 6th Edition, 1996 [3] GAO RFID Inc., “RFID Enabled Automated Parking Access Control Systems”, Retrieved 10. CONCLUSION May 5th, 2009 http://parking.gaorfid.com/

After accessing the performance of the proposed system, [4] Ronald, J. T. and Neal S. W., Digital System we were quite pleased with the results and the overall Principle and Application 7th Edition, Prentice – performance of the project. From the working of the IR Hall International Inc, London, 1998 remote, from the car to the receiver, gate mechanism [5] Hill, Frederick J. and Peterson Gerald R., driver and controls. The operation of the reset button at Digital Logic and Microprocessor, 1984. the control center and also the timing was excellent. [6] Vincent Tseng, Microprocessor Development and Development Systems [7] Shoewu, O. and O. Badejo. 2006, Radio Frequency Identification Technology: Development, Application and Security Issues, Pacific Journal of Science and Technology. 7(2): 144-152. [18] Jeffs Tyson(2008), http://www.howstuffswork.com

108

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

A Survey of Techniques For Answering Top-k queries

Neethu C V & Rejimol Robinson R R Dept. of Computer Science & Engineering SCT College of Engineering Trivandrum,India [email protected], [email protected]

ABSTRACT Top-k queries are useful in retrieving top-k records from a given set of records depending on the value of a function F on their attributes. Many techniques have been proposed in database literature for answering top-k queries.These are mainly categorized into three:Sorted-list based,layer based and View based. In first category, records are sorted along each dimension and then assigned a rank to each of the records using parallel scanning method.Threshold Algorithm(TA) and Fagin’s Algorithm(FA) are the examples of sorted-list based category. Second category is layer based category,in which all the records are organized into layers such as in onion technique and robust indexing technique.Third category includes methods such as PREFER and LPTA(Linear Programming Adaptation of Threshold Algorithm) and processing is based on the materialized views.

Keywords: Monotone functions,PREFER,Linearly optimally ordered set,Convex hull.

African Journal of Computing & ICT Reference Format Neethu C V & Rejimol Robinson R R (2013). A Survey of Techniques For Answering Top-k queries. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 109-116.

1. INTRODUCTION 1.2 Layer Based Category Top-k queries are intended for retrieving top-k records The algorithms in this category organize all records into from the database which are subjected to minimization or consecutive layers, such as Onion [4] and Robust maximization of the function F on the attributes of the Indexing Techniques [5]. The organization strategy is relation.This kind of queries appears frequently in many based on the common property among the records, such applications such as college ranking,job ranking etc.Due as the same convex hull layer in Onion [4]. Any top-k to the popularity of top-k queries, many techniques have query can be answered by up to k layers of records. The been proposed which are mainly includes sorted-list Onion indexing is based on a geometric property of based,layer based and view based techniques. convex hull, which guarantees that the optimal value can always be found at one or more of its vertices. 1.1 Sorted-list based Methods in this category sorts all records along each The Onion indexing makes use of this property to dimension and then assigned an overall grade to each of construct convex hulls in layers with outer layers the records based on the sorted lists.For example, enclosing inner layers geometrically. A data record is consider the example of college ranking.A student want indexed by its layer number or equivalently its depth in to join a college for doing graduation and he has some the layered convex hull. Queries with linear weightings preferences based on the attributes like distance to the issued at run time are evaluated from the outmost layer college,tution fee,university under which college is inwards. Onion indexing achieves orders of magnitude working,performance of the college for previous four speedup against sequential linear scan when N is small years etc.He then assigns grades to each of the attributes compared to the cardinality of the set. The Onion and sorted lists are created based on this assignment technique also enables progressive retrieval, which corresponding to each of the attributes.Then a list of processes and returns ranked results in a progressive colleges have retrieved based on their value for the query manner. Furthermore, the proposed indexing can be function.Here, the query function is a linear function in extended into a hierarchical organization of data to terms of the attributes of the records.FA and TA accommodate both global and local queries. [1],[2],[3] are the two techniques included in this category.

109

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Robust indexing [5] method is a kind layered technique PREFER is a layer on top of commercial relational for answering ranked queries. The layered indexing databases and allows the efficient evaluation of multi methods are less sensitive to the query weights. A key parametric ranked queries. LPTA[7] is a linear observation is that it may be beneficial to push a tuple as programming adaptation of the classical TA algorithm to deeply as possible so that it has less chance to be touched solve top-k query problem. in query execution. Motivated by this, a new criterion for sequentially layered indexing had been proposed: for any k, the number of tuples in top k layers is minimal in comparison with all the other layered alternatives. Since any top-k query can be answered by at most k layers, this proposal aims at minimizing the worst case performance on any top-k queries. Hence the proposed index is robust. While Onion and other layered techniques are sensitive to the query weights, This method, even though not optimal in some cases, has the best expected performance. Another appealing advantage of our proposal is that the top-k query processing can be seamlessly integrated into current commercial databases. Both Onion and other layered methods require the advanced query execution algorithms, which are not supported by many database query engines so far.

Figure 2.Example of top-k query processing.

2. TAXONOMY OF PROCESSING TOP-K QUERIES

Due to the high popularity of the top-k queries, various technique have been proposed for solving such situations.Supporting efficient top-k query processing in database system is relatively recent and active line of research. In the following subsection, all the important techniques included in above explained categoris have been explored in detail.

2.1 Naïve Algorithm To determine the top k objects, that is, k objects with the Figure 1.Classification of Top-k query evaluation highest overall grades, the naive algorithm must access techniques. every object in the database, to find its grade under each attribute. Steps of the Naïve algorithm[1] is given below. 1.3 View based category In view based techniques, the materialized views created  If (x1,x2,…,xm) are the grades of object R under from the relation can be used to answer top-k queries. the m attributes, then compute T(x1,x2,…,xm) PREFER[6] answers preference queries efficiently by overall grade of object R. using materialized views that have been preprocessed and stored.Queries with different weights will be first mapped  Sort the list of computed values. to the pre-computed order and then answered by determining the lower bound value on that order. When  Return top k rows corresponding to the sorted list. the query weights are close to the pre-computed weights, the query can be answered extremely fast. Unfortunately, The main disadvantage of the Naïve algorithm is the this method is very sensitive to weighting parameters. A large processing time when dealing with large databases. reasonable derivation of the query weights (from the pre- computed weights) may severely deteriorate the query performance.

110

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2.2 Fagin’s Algorithm Fagin introduced an algorithm (‘‘Fagin’s Algorithm[1]’’,  Then compute the grade t(R) =t(x1,x2 ,…xm) of or FA), which often does much better than the naive object R: If this grade is one of the k highest we algorithm. In the case where the orderings in the sorted have seen, then remember object R and its grade lists are probabilistically independent, FA finds the top k t(R). answers, over a database with N objects with arbitrarily high probability. This algorithm is implemented in  For each list Li, let xi be the grade of the last Garlic, an experimental IBM middleware system. object seen under sorted access. Define the threshold value ψ to be t(x1,x2,….,xm). As soon  Do sorted access in parallel to each of the m sorted as at least k objects have been seen whose grade lists Li: Wait until there are at least k ‘‘matches’’, is at least equal to ψ then halt. that is, wait until there is a set of at least k objects such that each of these objects has been seen in each  Let Y be a set containing the k objects that have of the m lists. been seen with the highest grades. The output is then the graded set {(R, t(R)) | R€Y}.  For each object R that has been seen, do random access as needed to each of the lists Li to find the ith The algorithm scans multiple lists, representing different field xi of R: rankings of the same set of objects. An upper bound T is maintained for the overall score of unseen objects. The  Compute the grade t(R)= t(x1,x2,….xm) for each upper bound is computed by applying the scoring object R that has been seen. Let Y be a set function to the partial scores of the last seen objects in containing the k objects that have been seen with the different lists. Notice that the last seen objects in different highest grades (ties are broken arbitrarily). The lists could be different. The upper bound is updated every output is then the graded set {(R, t(R)) | R€Y}. time a new object appears in one of the lists. The overall score of some seen object is computed by applying the Fagin shows that his algorithm is optimal with high scoring function to object’s partial scores, obtained from probability in the worst case if the aggregation function is different lists. To obtain such partial scores, each newly strict (so that, intuitively, we are dealing with a notion of seen object in one of the lists is looked up in all other conjunction),and if the orderings in the sorted lists are lists, and its scores are aggregated using the scoring probabilistically independent. In fact, the access pattern function to obtain the overall score. All objects with total of FA is oblivious to the choice of aggregation function, scores that are greater than or equal to T can be reported. and so for each fixed database, the middleware cost of The algorithm terminates after returning the kth output. FA is exactly the same no matter what the aggregation Example 1 given below illustrates the processing of TA. function is. This is true even for a constant aggregation function; in this case, of course, there is a trivial Example 1[10]:Consider two data sources containing algorithm that gives us the top k answers (any k objects same set of objects.Let A1 and A2 are the attributes in will do) with O(1) middleware cost. two data sources respectively.The Query function,F is defined as F=A1+10*A2. The working of TA is depicted So FA is not optimal in any sense for some monotone in the following figure. aggregation functions t: As a more interesting example, when the aggregation function is max (which is not strict), it is shown in that there is a simple algorithm that makes at most m*k sorted accesses and no random accesses that finds the top k answers. By contrast, the algorithm TA is instance optimal for every monotone aggregation function, under very weak assumptions.

2.3 Threshold Algorithm Even in the cases where FA is optimal, this optimality holds only in the worst case, with high probability. This leaves open the possibility that there are some algorithms that have much better middleware cost than FA over certain databases. The algorithm TA, which we now discuss, is such an algorithm.  Do sorted access in parallel to each of the m sorted lists Li: As an object R is seen under sorted access in some list, do random access to the other lists to find the grade xi of object R in

every list Li.

111

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

In the first step, retrieving the top object from each list, and probing the value of its other attribute value in the other list, result in revealing the exact scores for the top objects. The seen objects are buffered in the order of their scores. A threshold value, T, for the scores of unseen objects is computed by applying F to the last seen scores in both lists, which results in 70+6*10=130. Since both seen objects have scores less than T, no results can be reported. In the second step, T drops to 90, and objects 4 and 2 can be safely reported since its score is above T. The algorithm continues until k objects are reported, or sources are exhausted.

2.4 Onion Technique Procedure for index creation: This technique comes under the layer based category and uses a special indexing structure for answering top-k Step1:Input a set of records R and iterate the queries. The Onion indexing is based on a geometric following steps until size(R) becomes less than property of convex hull, which guarantees that the zero. optimal value can always be found at one or more of its vertices. The Onion indexing makes use of this property Step 2:Construct convex hull of the data records R. to construct convex hulls in layers with outer layers enclosing inner layers geometrically. A data record is Step 3:Store the records of hull vertices in set Vi. indexed by its layer number or equivalently its depth in the layered convex hull. Queries with linear weightings Step4:Assign records in set V to layer k. issued at run time are evaluated from the outmost layer inwards. Basic idea of the onion technique is that Step 5:Set R=R-V and k=k+1. partition the collection of d-dimensional data points into sets that are optimally linearly ordered. This property is used to construct convex hulls in layers with outer layers enclosing inner layers geometrically.

Definition 1.Optimally Linearly Ordered Set:A collection of sets{s1,s2,…,sn}are optimally linearly ordered sets if and only if a d-dimensional vector ā,

Ǝ ō ϵ si such that

t t t for every ĉ ϵ si+j ,j>0, ā ō> ā ĉ where ā ō represents the inner product of two vectors.

Partitioning a set of data points into optimally linearly ordered sets is based on the following theorem. This indexing structure can be used for query Theorem 1: Given a set of records R mapped to a d- evaluation.Onion indexing achieves orders of magnitude dimensional space, and a linear maximization criterion, speedup against sequential linear scan when N is small the maximum objective value is achieved at one or more compared to the cardinality of the set. The Onion vertices of the convex hull of R. technique also enables progressive retrieval, which processes and returns ranked results in a progressive Definition 2. A set S is convex if whenever two points P manner. Furthermore, the proposed indexing can be and Q are inside S, then the whole line segment PQ is extended into a hierarchical organization of data to also in S. accommodate both global and local queries.

112

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2.5 Robust Indexing Structure The properties have attributes such as price, number of This is an another layered indexing structure useful for bedrooms, age, square feet, etc. For a user, the price of a the evaluation of top-k queries. The idea of multi-layer property and the square feet area may be the most indexing has been also adopted by to provide robust important issues, equally weighted in the final choice of a indexing[5],[10] for top-k queries. Robustness is defined property, and the property’s age may also be an important in terms of providing the best possible performance in issue, but of lesser weight. The vast majority of e- worst case scenario, which is fully scanning the first k commerce systems available for such applications do not layers to find the top-k answers. The main idea is that if help users in answering such queries, as they commonly each object Oi is pushed to the deepest possible layer, its order according to a single attribute. In these cases, retrieval can be avoided if it is unnecessary. This is preference queries have significant role and for PRFER accomplished by searching for the minimum rank of each system also. object oi in all linear scoring functions. Such rank represents the layer number, denoted l*(Oi), where object 4. LPTA Oi is pushed to. For n objects having d scoring predicates, computing the exact layer numbers for all Algorithm(LPTA)[7],[10] is another technique included objects has a complexity of O(nd log n), which is an in the view based category.It performs much better than overkill when n or d are large. Approximation is used to PREFER. reduce the computation cost. An approximate layer number, denoted l(Oi), is computedsuch that l(Oi) · Problem 1: (Top-K Query Answer Using Views). Given l*(Oi), which ensures that no false positives are produced a set U of views, and a query Q, obtain an answer to Q in the top-k query answer. combining all the information conveyed by the views in U.

3. PREFER Consider a single relation R with m numeric attributes X1,X2,….Xm, and n tuples t1, . . . , tn. Let Domi = [lbi, This is a view based evaluation of the top-k queries. ubi] be the domain of the ith attribute. Refer to table R as Recent successful work in non-layered approaches a base table. Each tuple t may be viewed as a numeric includes the PREFER system [6],[10], where tuples are vector t = (t[1], t[2], . . . , t[m]). Each tuple is associated sorted by a pre-computed linear weighting configuration with a tuple-id (tid).Here consider top-k ranking queries, Users often need to optimize the selection of objects by which can be expressed in SQL-like syntax: SELECT appropriately weighting the importance of multiple object TOP [k] FROM R WHERE RangeQ ORDER BY attributes. Such optimization problems appear often in ScoreQ. More abstractly, a ranking query may be operations research and applied mathematics as well as expressed as a triple Q = (ScoreQ, k, RangeQ), where everyday life; e.g., a buyer may select a home as a ScoreQ(t) is a function that assigns a numeric score to any weighted function of a number of attributes like its tuple t (the function does not necessarily involve all distance from office, its price, its area, etc. attributes of the table), and RangeQ(t) is a Boolean function that defines a selection condition for the tuples The queries here use a weight function over a relation’s of R in the form of a conjunction of range restrictions on attributes to derive a score for each tuple. Database Domi, i 2 {1, . . . ,m}. Each range restriction is of the systems cannot efficiently produce the top results of a form li ≤ Xi ≤ ui, I ϵ {1, . . . ,m} and the interval [li, ui] preference query because they need to evaluate the Domi.The semantics requires that the system retrieve weight function over all tuples of the relation. the k tuples with the top scores satisfying the selection PREFER[6] answers preference queries efficiently by condition. using materialized views that have been preprocessed and stored.Queries with different weights will be first mapped LPTA[7] is a linear programming adaptation of the to the pre-computed order and then answered by classical TA algorithm to solve Problem 1.1 for the determining the lower bound value on that order. When special case when views and queries are of the form V 0 the query weights are close to the pre-computed weights, = (ScoreV 0 , n, *) and Q = (ScoreQ, k, *) respectively. the query can be answered extremely fast. Unfortunately, Consider a relation with attributes X1, X2 and X3 as this method is very sensitive to weighting parameters. A shown in Figure 1.3.2.1. Let views V1 and V2 have reasonable derivation of the query weights (from the pre- scoring functions f1, f2 respectively as shown in Figure computed weights) may severely deteriorate the query 1.3.2.1 and consider a query Q = (f3, k, *). The algorithm performance. PREFER is a layer on top of commercial initializes the top-k buffer to empty. It then starts relational databases and allows the efficient evaluation retrieving the tids from the views V1, V2 in a lock-step of multi parametric ranked queries For example consider fashion, in the order of decreasing score (w.r.t. the view’s a database containing houses available for sale. scoring functions).

113

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

For each tid read, the algorithm retrieves the The top-2 buffer remains unchanged and the linear 1 2 corresponding tuple by random access on R, computes its program is solved one more time using sd = 299 and sd score according to the query’s scoring function f3, = 202. This time, unseenmax= 953.5 < topkmax = 1248 and updates the top-k buffer to contain the top-k largest the algorithm terminates. Thus, in total LPTA conducts scores (according to the query’s scoring function), and two sequential and two random accesses per view. In checks for the stopping condition as follows: After the contrast, the TA algorithm executed on R of Figure 1 will dth iteration, let the last tuple read from view V1 be identify the correct top-2 results after 12 sorted and 12 1 1 2 2 (tidd , sd ) and from view V2 be (tidd , sd ). Let the random accesses in total. The performance advantage of minimum score in the top-k buffer be topkmin. At this LPTA is evident. stage, the unseen tuples in the view have to satisfy the following inequalities (the domain of each attribute of R 5. ALGORITHM LPTA(U,Q) of Figure is [1, 100]). U={V1,…,Vr}//set of views Q=(ScoreQ ,k,*)//Query Topk-Buffer={} topkmin =0 for d=1 to n do for all views Vi(1≤i≤r) in block-step do i i Let(tid d, s d ) be the d-th item in Vi //Update top-k buffer i i Let t d =RandomAccess(tid d) i if ScoreQ(t d)>topkmin then if(|topk-Buffer|=k) then Remove min score tuple from topk-Buffer end if i i Add(tid d,ScoreQ(tid d)) to topk-buffer Topkmin=min score of topk-Buffer end if // Checking stopping conditions by solving LP Figure 3.Example of views. Let Unseen =convex region defined by max lbj≤Xj≤ubj for every 1≤j≤m The following system of inequalities defines a j Scorevj ≤s d for every 1≤j≤r convex region in three dimensional space. Compute Unseen =max {Score (t)} max tϵunseen Q If(|topk-Buffer|=k)and (Unseenmax≤topkmin) Then Return topk-Buffer end if end for end for This system of inequalities defines a convex region in three dimensional space. Let unseenmax be the solution to 6. COMPARISON OF DIFFERENT TECHNIQUES the linear program where we maximize the function f3 = 3X1 + 10X2 + 5X3 subject to these inequalities. It is easy This section includes comparison of different techniques to see that unseenmax represents the maximum possible employed in the top-k query evaluation. The comparison score (with respect to the ranking query’s scoring is performed based on the three important criteria which function) of any tuple not yet visited in the views. The are ranking function, ranking model and data access algorithm terminates when the top-k buffer is full and operation involved in the different techniques. The unseenmax ≤ topkmin. Considering the example of given ranking function can be generic or monotone. Most of the figure, the algorithm will proceed as follows; current top processing techniques assume monotone ranking functions since they fit in many practical First retrieve tid and conduct a random access to R to scenarios, and have appealing properties allowing for retrieve the full tuple and tid 6 from V2 accessing R efficient top-k processing. But Few recent techniques i again. The top-2 buffer contains the following pairs (tidd , address top-k queries in the context of constrained i sd ) {(7, 1248), (6, 996)}. The solution to the linear function optimization. The ranking function in this case is program with s1q= 527 and s2d = 219 yields an allowed to take a generic form. unseenmax =1338 > topkmax = 1248 and the algorithm conducts one more iteration.This time we access tid 6 from V1 and tid 4 from V2.

114

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFERENCES

[1] R. Fagin, A. Lotem, and M. Naor, “Optimal Aggregation Algorithms for Middleware,” Proc. Symp. Principles of Database Systems (PODS), 2001.

[2] S. Nepal and M.V. Ramakrishna, “Query Processing Issues in Image (Multimedia) Databases,” Proc. 15th Int’l Conf. Data Eng. (ICDE), 1999.

[3] U. Guntzer,W.T.Balke, and W. Kiebling, “Optimizing Multi-Feature Queries for Image Databases,” Proc. Int’l Conf. Very Large Data Bases (VLDB), 2000.

[4] Y.-C. Chang, L.D. Bergman, V. Castelli, C.S. Li, M.L. Lo, and J.R. Smith, “The Onion Technique:

Another criteria is ranking model. It can be top-k join or Indexing for Linear Optimization Queries,” Proc. ACM top-k selection. In top-k selection model, the scores are SIGMOD, 2000. assumed to be attached to base tuples. A top-k selection query is required to report the k tuples with the highest [5] D. Xin,C.Chen, and J. Han, “Towards Robust scores. Scores might not be readily available since they Indexing for Ranked Queries,” Proc. Int’l Conf. Very could be the outcome of some user-de Consider a set of Large Data Bases (VLDB),2006. relations R1 ,….,Rn. A top-k join query joins R1,…,Rn, and returns the k join results with the largest combined [6] V. Hristidis, N. Koudas, and Y. Papakonstantinou, scores. The combined score of each join result is “Prefer: A System for the Efficient Execution of Multi- computed according to some function F(p1,…., pm), Parametric Ranked Queries,” Proc. ACM SIGMOD, where p1,….,pm are scoring predicates defined over the 2001. join results.fined scoring function that aggregates information coming from different tuple attributes. Third [7] G. Das, D. Gunopulos, N. Koudas, and D. criteria is data access which can be sorted access or Tsirogiannis, “Answering Top-K Queries Using Views,” random access.In sorted access, Object R has the lth Proc. Int’l Conf. Very Large Data Bases (VLDB), 2006. highest grade in the ith list, then l sorted accesses to the ith list are required to see the grade under sorted access [8] S. Bo¨ rzso¨ nyi, D. Kossmann, and K. Stocker, “The and in random access, grade of object R in the ith list Skyline Operator,” Proc. 17th Int’l Conf. Data Eng. obtains it in one random access. (ICDE), 2001.

7. CONCLUSION [9] D. Papadias, Y. Tao, G. Fu, and B. Seeger, “An Optimal and Progressive Algorithm for Skyline Queries,” A surevey of top-k query processing techniques based on Proc. ACM SIGMOD,2003. the different criterias have done.For this purpose, a detailed analysis of different techniques included in three [10] Ihab F. Ilyas, George Beskales And Mohamed A. important categories like sorted-list based category,layer Soliman,”A survey of top-k query processing technique based category and view based category have explored. in relational database systems” University of Waterloo, Support was provided in part by the Natural Sciences and Engineering Research Council of Canada 2011.

[11] D. Kossmann, F. Ramsak, and S. Rost, “Shooting Stars in the Sky:An Online Algorithm for Skyline Queries,” Proc. Int’l Conf. Very Large Data Bases (VLDB), 2002.

[12] T.H. Cormen, C.E. Leiserson, R.L. Rivest, and C. Stein, Introduction to Algorithms. MIT Press, 2001.

[13] D. Campbell and R. Nagahisa, “A Foundation for Pareto Aggregation,” J. Economical Theory, vol. 64, pp. 277-285, 1994.

115

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

[14] M. Voorneveld, “Characterization of Pareto Dominance,” Operations Research Letters, vol. 32, no. 3, pp. 7-11, 2003.

[15] C. Li, B.C. Ooi, A.K.H. Tung, and S. Wang, “DADA: A Data Cube for Dominant Relationship Analysis,” Proc. ACM SIGMOD, 2006.

[16] Y. Tao, V. Hristidis, D. Papadias, and Y. Papakonstantinou, “Branch-and-Bound Processing of Ranked Queries,” Information Systems, vol. 32, no. 3, pp. 424-445, 2007.

[17] R.J. Lipton, J.F. Naughton, and D.A. Schneider, “Practical Selectivity Estimation through Adaptive Sampling,” Proc. ACM SIGMOD, 1990.

[18] R.J. Lipton and J.F. Naughton, “Query Size Estimation by Adaptive Sampling,” Proc. Symp. Principles of Database Systems(PODS), 1990.

[19] S. Chaudhuri, N.N. Dalvi, and R. Kaushik, “Robust Cardinality and Cost Estimation for Skyline Operator,” Proc. 22nd Int’l Conf. Data Eng. (ICDE), 2006.

[20] C.B. Barber, D.P. Dobkin, and H. Huhdanpaa, “The QuickhullAlgorithm for Convex Hulls,” ACM Trans. Math. Software, vol. 22, pp. 469-483, 1996.

[21] D. Xin, J. Han, H. Cheng, and X. Li, “Answering Top-k Queries with Multi-Dimensional Selections: The Ranking Cube Approach,” Proc. Int’l Conf. Very Large Data Bases (VLDB), 2006.

[22] S. Nepal and M. Ramakrishna, “Query Processing Issues inImage(Multimedia) Databases,” Proc. 15th Int’l Conf. Data Eng. (ICDE), 1999.

[23] M. Li and Y. Liu, “Iso-Map: Energy-Efficient Contour Mapping in Wireless Sensor Networks,” IEEE Trans. Knowledge and Data Eng., vol. 22, no. 5, pp. 699- 710, May 2010.

[24] H. Bast, D. Majumdar, R. Schenkel, M. Theobald, and G. Weikum, “IO-Top-k: Index-Access Optimized Top-k Query Processing,” Proc. Int’l Conf. Very Large Data Bases (VLDB), 2006.

[25] N. Mamoulis, K.H. Cheng, M.L. Yiu, and D.W. Cheung, “Efficient Aggregation of Ranked Inputs,” Proc. 22nd Int’l Conf. Data Eng.(ICDE), 2006.

116

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Performance Evaluation of AN-VE: An Error Detection and Correction Code

Egwali Annie O. and Akwukwuma V. V. N. Department of Computer Science Faculty of Physical Sciences. University of Benin. P.M.B. 1154. Benin City. Nigeria. [email protected]; [email protected]

ABSTRACT Several techniques have been proposed and employed to effectively detect and correct errors introduced during message transmission over a communication channel or at the destination domain during storage. Some of these techniques can detect: only single error, all unidirectional errors, only burst errors, any number of zero-to-one bit-flip errors as long as no one-to-zero bit-flip errors occur in the same stream of message bits, errors with known positions, assume a code is correct if the error positions are known or cannot detect errors which appear in the same position in a pair of message codes. Coding techniques that detects and correct errors are more precise at detecting error positions and correcting them, however if more than one error occur, it becomes a challenge to detect all errors and decode correctly. In this paper, the performance of AN-VE is evaluated in comparison with the following coding techniques: Reed-Solomon codes, Hamming codes, and Low-density parity-check codes. An independent design platform is utilized for the implementation via Simulink, which shows a significant reduction in uncorrected errors during message transmission. The efficient performance of AN-VE makes it a more applicable coding technique for telecommunication, data compression and other application.

Keywords: Burse error, Hamming code, Message bits, Low-Density Parity-Check, Reed-Solomon codes. African Journal of Computing & ICT Reference Format Egwali Annie O. & Akwukwuma V. V. N. (2013). Performance Evaluation of AN-VE: An Error Detection and Correction Code. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 117-126.

1. INTRODUCTION

Components of communication systems like computer These errors can result in an original messages being memory, satellite, magnetic and optical data storage transmitted over a communication channel to become media, network communications and space erroneous at the destination domain and may result in communications can be used to perform various intermediate routers making decisions based on bogus operations on information such as storage (using data and misdirect packets. There are basically two types registers, RAM etc.), transmission (interconnects, lines of erroneous data that can be generated at the etc.) and processing (using CPU, logic elements etc). But communication line. These are single-bit error and burst environmental interference, signal distortion or error. In single-bit error only one bit of a word given attenuation (sender and receiver out of synchronization), stream of message bits (SMB) is changed from 1 to 0 or thermal noise that can randomly corrupt bits, physical from 0 to 1 as shown in figure 1. Single bit errors are defects in the communication medium, channel noise that more likely to occur in a parallel transmission than in a changes a 0 level into a 1 level, ignored end-to-end serial transmission. In a parallel transmission were 16 principles by transport protocols and user application wires may be used to send a 16 SMB at the same time, if programs, system malfunction, storage error (DRAM one of the wires is noisy, one bit in the SMB will be memory cell contents being changed spuriously due to corrupted. electromagnetic interference, magnetic flux density increase in magnetic storage devices resulting in one or more bit flip etc) and components failures can invert transmitted codes and cause random bit errors.

117

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

(Sent data) 0 1 0 0 1 0 1 1 1 0 1 1 0 1 0 1 Erroneous single bit change of 0 to 1

0 1 1 0 1 0 1 1 1 0 1 1 0 1 0 1 (Received data)

Figur e 1: Single bit error

The term burst error means that two or more bits in the SMB change from 0 to 1 or vice-versa. Unlike single-bit error, burst errors are likely to occur in serial transmission mode. Burst errors are highly correlated for if one bit has an error, it is likely that the adjacent and intermediate bits could be corrupted also (see figure 2).

8 bits length of burst error (sent data) 0 1 0 0 1 0 1 1 1 0 1 1 0 1 1 0

0 1 1 1 1 0 1 1 0 1 1 1 0 1 1 0

Erroneous bits (received data)

Figure 2: Burst Error

Error detecting and correcting models are more thorough 2. PREVIOUS WORKS by introducing extra redundant codes to detect the actual position of errors and correcting them, however for some Forney (1966) revealed that when RS codes are serially models if more than one error occurs, it becomes difficult concatenated with convolutional codes, RS codes to detect all errors and decode correctly. On the other weakness against burst errors are counteracted. Plummer hand there are some models for which the full-duplex (1989) and Sheinwald et al, (2002) presented an analytic mode is not permitted with the presence of errors, for comparison of checksum coding techniques. McAuley such system, there is no likelihood of a message (1994) investigated Fletcher checksum and CRC coding retransmission, and so the receiver has to implement methods and proposed the Weighted Sum Codes (WSC) some error-correction algorithm to accurately decode the algorithm as an alternative method. Feldmeier (1995) message. In an initial work (Egwali and Akwukwuma, carried out an analytical evaluation of weighted sum code 2011), an error detecting and correcting code model, AN- alongside Fletcher checksum, XOR checksum, one’s VE, was presented that simultaneously: verifies the complement addition checksum, CRC and block parity. length of the encoded message, analyzes and detects error He asserted that WSC has high computational speed as codes from the encoded message even if similar error Fletcher checksum and error detection capability as CRC. occurs within the same position on all clusters of bits in Berrou et al, (1995) discovered a class of codes termed the encoded SMB, addresses all error positions via the turbo codes that exhibit near Shannon limit performance use of extra parity bits and corrects all errors accurately by means of iterative decoding algorithms. Partridge et even if more than one errors occur in a decoded SMB and al, (1995), Stone et al, (1998) and Stone and Partridge finally perform message verification at the receiver (2000) explored checksum efficacy of Fletcher domain. In this paper, the performance of AN-VE is checksum, CRC and one’s complement addition evaluated in comparison with the following coding checksum. The study reveals that Fletcher checksum and techniques: Reed-Solomon codes, Hamming codes, and one’s complement addition checksum have great Low-density parity-check codes. An independent design probabilities of undetected errors due to non-uniformity platform is utilized for the implementation via Simulink, of network data. which shows a significant reduction in uncorrected errors during message transmission.

118

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Baicheva et al, (1998; 2000) and Kazakov (2001) - Position 32: check 32 bits, skip 32 bits, examined the efficiency of optimal CRC polynomials for check 32 bits, skip 32 bits, etc. (32-63,96- specific lengths. Richardson et al, (2001) analysis on 127, etc.) coding methods revealed that bit error performance - Position 64: check 64 bits, skip 64 bits, deteriorates as the data bits length decreases in size. check 64 bits, skip 64 bits, etc. (64-127, Koopman (2002) and Koopman and Chakravarty (2004) 192-255, etc.) investigated CRC polynomials and proposed a - etc. polynomial selection process for embedded networks. . Step 4: Set a parity bit to 0 if the total number of Egwali and Akwukwuma (2011) presented a hybrid error ones in the positions it checks is even. Set a parity detecting technique, AN-VE, which simultaneously bit to 1 if the total number of ones in the positions it detect the existence of faulted codes right from the checks is odd. transmitter domain and analyzes all error positions in the encoded message. Sometimes an original SMB can be intersected by noise 3. ANALYSIS OF LINEAR BLOCK CODES in the channel, using varied specified error probabilities, Generally, computer systems process information in which can switch some of the 0(s) and 1(s) in the original blocks of bits and corpus of block codes can be separated SMB. Hence an analysis of the checks bits will reveal into linear and systematic codes. Linear codes, which is the check bits that are affected. The key to the Hamming the focus of this paper involves the linear combination of coding technique is the use of extra parity bits that allow valid code words to produce a new valid code word. the detection of just a single error. Repetition coding Linear code length is express in terms of the number of technique involves repeating a SMB which is divided blocks n and the number of bits k contained in each into blocks of bits, across a channel two or three times to block. The following are examples of linear block codes: achieve error-free communication. Every SMB block is transmitted at some specified number of times. While Hamming (1980) posited the hamming error and repetition coding techniques are uncomplicated and find correcting coding technique which are the earliest linear applications for both error correction and detection error correcting code technique. It involves the use of an implementation, they are very ineffective, and can be extra parity bit to ensure the identification of a single vulnerable to difficulties if the error occurs in the exact error. Generally, a Hamming (f, g) code consists of “g” same position in each of the transmitted SMB. number of information bits and the derived encoded data bits of length f. “s” denotes the check bits such that g = Low-Density Parity-Check (LDPC) codes are linear error 2s − s − 1, f = 2s − 1, s 3 (Hamming, 1980). Hence for detection and correction codes with long block lengths and sparse parity-check matrices. It is composed of an a given byte of data, to create the SMB, the following encoder and a decoder that decodes generic binary LDPC sequences of steps are executed: codes iteratively, execute a user-specified number of

iterations until all parity-checks are satisfied and output . Step 1: Create the SMB as parity bit positions that hard or soft decisions for decoded bits. LDPC are are powers of two (i.e. positions 1, 2, 4, 8, 16, etc.). designed in such a way that all bits act equivalently. Each . Step 2: Next all other bit positions are used to parity check bit s checks some small fixed g є f bits and encode the data to be encoded. (i.e. positions 3, 5, 6, each bit is checked by some small fixed j є t parity check 7, 9, 10, 11, 12, 13, 14, 15, 17, etc.). bits. The Reed-Solomon codes (Reed-Solomon, 2011) is . Step 3: Use each parity bit position of step 1 is used a linear cyclic systematic non-binary block code (i.e. the to calculate the parity for some of the bits in the transmitted message is partitioned into separate blocks of SMB. The position of the parity bit determines the data and each block of data has parity protection sequence of bits that it alternately checks and skips. information appended to it to form an independent code Thus parity bit at: word). Reed-Solomon Codes are useful for correcting - Position 1: check 1 bit, skip 1 bit, check 1 errors that occur in bursts. In the simplest case, it is bit, skip 1 bit, etc. (1,3,5,7,9,11, etc) denoted as RS(f, g) where f is the data bit length of total - Position 2: check 2 bits, skip 2 bits, check data bits and where g is the number of information bits 2 bits, skip 2 bits, etc. (2,3,6,7,10,11, etc.) and is of the form N= 2M-1. 2M is the number of - Position 4: check 4 bits, skip 4 bits, check symbols for the code (Clarke, 2002). 4 bits, skip 4 bits, etc.

(4,5,6,7,12,13,14,15, etc.)

- Position 8: check 8 bits, skip 8 bits, check

8 bits, skip 8 bits, etc. (8-15,24-31, etc.)

- Position 16: check 16 bits, skip 16 bits,

check 16 bits, skip 16 bits, etc. (16-31,48-

63, etc.)

119

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The error-correction capability of a Reed-Solomon code 4.1 AN-VE Error Detecting Mode is floor ((N-K)/2), where floor rounds the elements of AN-VE error detecting mode involves the ((N-K)/2) to the nearest integers less than or equal to ((N- following five steps: K)/2. g is the data bit length of the message words. The Step 1: Generate the SMB as parity bit positions difference N-K must be even. It is sometimes convenient that are powers of two (i.e. positions 1, 2, to use a shortened Reed-Solomon code in which N is less 4, 8, 16, 32 etc.). than 2M-1. In this case, the encoder appends 2M-1-N Step 2: All other bit positions are used to encode the zero symbols to each message word and data bits. The data to be encoded. (i.e. positions 3, 5, 6, systematic code (i.e. the encoding process does not alter 7, 9, 10, 11, 12, 13, 15, etc.). the bits of the data bits and parity codes are added as a Step 3: Each parity bit position of step 1 is used to separate part of the block). Figure 3 represents a typical calculate the parity for some of the bits in Reed-Solomon data bits. In the Reed-Solomon data bits, the SMB. The position of the parity bit g symbols are unchanged and the 2t parity symbols are determines the sequence of bits that it appended at the right end to make an f block length data alternately checked and skipped. bits symbol (where 2t = f - g). Step 4: A parity bit is set to 0 if the total number of ones in the positions it checks is even. Else the parity bit is set to 1 if the total number of ones in the positions it checks f is odd. Step 5: If the created SMB does not match the g 2t received SMB, a verification of the received SMB code block error bits DATA PARITY positions is established. Step 6: Inputted bit in the received SMB is checked for accidental changes by lining Figure 3: Reed Solomon Data bits input bits in a row, and a (n+1)-bit check divisor is positioned underneath the left- In the encoder, redundant bits are generated using a hand end of the SMB row. If the input bit generator polynomial and appended to the block length of above the leftmost divisor bit is 0, the bit total message data bits. In the decoder error location and is left and the divisor is moved to the right size are calculated using the generator polynomial which by one bit. If the input bit above the can correct up to t bits that contain errors in the data bits leftmost divisor bit is 1, the divisor is of the received message. The decoders also indicate how XORed into the input. The divisor is then many errors detected while decoding. The RS code is shifted one bit to the right, and the process very efficient at addressing bursts errors (i.e. uses is repeated until the divisor reaches the symbols and not bits) because, although a symbol may right-hand end of the input row. Since the have all its bits in error, this counts as only one symbol leftmost divisor bit zeroed every input bit error in terms of the correction capacity of the code. RS it touched, when this process ends the only codes are an exceptional solution for multicast because it bits in the input row that can be nonzero does not require a back-channel. are the n bits at the right-hand end of the row, which will always be less than the 4. AN-VE Detecting and Correcting Technique divisor. AN-VE (f, g) code consists of “g” data bits and the Step 7: After checking for changes, if the created encoded data bits of length f. g is defined by the SMB does not match the received SMB, equation: g = 2a – a − 1, and f is defined by the equation, the system establishes that the received f = 2a − 1, were “a” denotes the parity bits such that a code block contains data error and take 3. To ensure that the original SMB is not intersected by corrective measures to detect the actual bit noise in the channel, AN-VE employs varied error locations containing the errors. probabilities, which switches some of the 0(s) and 1(s) in the original SMB.

120

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4.2 AN-VE Error Correcting Mode original data bits are inputted via the communication Step 8: The affected check bits positions are source through the channel, and finally the decoded data established. bits is displayed at the output sink. Basically, in every Step 9: To effectively correct all errors, AN-VE communication system the basic components of employs a parsing procedure that parses communication used to simulate the aforementioned each n-bit binary position on the received coding system includes the signal (original data bits), a SMB which are lined in a row and source for the signal or original data bits, a channel with compares it with the n-bit binary position noise, error rate calculation block, the output display of the created SMB which are lined in a block and the decoded data bits. row starting from the extreme left. The - parity check bits of the first bit at variance Hamming Coding Technique: between the two sets of n-bit binary The logical inference and relationship between the position in a row are verified to detect the Hamming coding communication components are error, which is then corrected at each represented in figure 4. The Bernoulli Binary Generator parsing stage (the system only block is the source x for the signal in this model. It acknowledges the positions of the other generates a random binary sequence of numbers using a bits in the row and not their values). Bernoulli distribution. The Bernoulli distribution with Step 10: Repeat step 9 till all n-bit binary positions parameter p produces zero with probability p and one is parsed and both created and received with probability 1-p. The Bernoulli distribution has mean codes are equivalent. value 1-p and variance p(1-p). The Probability of a zero parameter specifies p, and can be any real number 5. EXPERIMENTATION between zero and one. Subsequently the Hamming encoder encodes the original number of information bits Analytical experimentation was carried out using Matlab g before it is sent through the channel. It creates a 7.1.0.246 (R14) Service Pack 3 with Simulink 6.3 for the Hamming code with message length g and data bit length simulations of the performance of AN-VE, Hamming f. The number f must have the form 2M-1, where M is an codes and Low-density parity-check codes in the integer greater than or equal to 3. Also g must equals f-M. presence of noise. To execute each coding operation, the

Figure 4: Simulink of the Hamming Coding Communication Components (readings of Hamming (1023, 1013))

121

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The Binary Symmetric Channel (BSC) simulates a displays the number of errors occurrences. The third box channel with noise. The channel generates a random of sk displays the total number of bits (bn) transmitted. binary signal, and then switches the symbols 0 and 1 or the reverse in the signal, according to a specified error Low-Density Parity-Check (LDPC): probability, to simulate a channel with noise. The The Simulink of the Low-Density Parity-Check Hamming decoder decodes the data once it is launch communication components are represented in figure 5. through the channel and verifies if an error is created in g The Bernoulli Binary Generator is also the source x for by the noise in the channel. It identifies the error and the signal. The BCH Encoder block creates a BCH code decodes the received data back to g correctly. with data bit length g and data bit length f. For a given The input must contain exactly f elements and the output message data bit length f, only specific data bits g are is a vector of length g. The bits error rate represented as valid for a BCH code. For a full length BCH code, f must err is computed at the Error Rate Calculation block be of the form 2M-1. If f is less than 2M-1, the block denoted as z, which in this case detects and computes the assumes that the code has been shortened by length 2M - error rate of the channel using the values of its two input 1 - f. However, if f is greater than or equal to 2M-1, ports, the transmitted signal Tx and the received signal Primitive polynomial must be specified to appropriately Rx. The computation is based on the following equation set the value of M. The LDPC encoder block, encodes (1), which denotes the probability of two or more errors the binary low-density parity-check code specified by occurring in encoded data bits of length f. parity-check matrix with dimension f – g by f (where n >k > 0) of real numbers. All nonzero elements are equal to 1. err = 1 – (0.99)f – f(0.99)f- The upper bound limit for the value of n is 2(31)-1. Both 1(0.01)………………………..equ. (1) the input and output are discrete-time signal, the output inherits the data type of the input, and the input must be The bits error rate result is displayed in the first box of sk binary-valued (0 or 1). after the termination of execution. The second box of sk

General BBFRAME QPSK Bernoulli Block Buffering BCH Encoder LDPC Encoder Modulator Binary Interleaver x

Tx PER Tx BER Tx Error Rate PER #Errs #Errs AWGN SNRdB Calculation Rx Rx Rx #Pkts #Bits SNR (dB)

Packet Error Rate z-1 sk-1

BBFRAME Out Unbuffering BCH Decoder General QPSK Iter Block Soft-Decision LDPC Decoder Deinterleaver Demodulator ParChk RX Constellation

No. of Parity-Check Failures

No. of Iterations

Figure 5: Simulink of the Low-Density Parity-Check Communication Components.

122

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The General Block Interleaver block rearranges the AN-VE: elements of its input vector without repeating or omitting For AN-VE, the source x is the Bernoulli Binary any elements. If the message contains f data bit length of Generator. Next the Hamming encoder encodes the bits, then the bits parameter is a vector of length f that original stream of message bits g before it is sent through indicates the indices, in order, of the input bits that form the Binary Symmetric Channel (BSC) that adds binary the length-f output vector; that is, errors to the input signal (see figure 6). The cyclic encoder creates the systematic cyclic bits with message Output(g) = Input(bits(g)) length K and encoded SMB length N in the binary cyclic encoder. The code rate is: for each integer g between 1 and f. The data bit length must be integers between 1 and f, and must have no Code rate = ge/fe = message bits length/ encoded SMB repetitions. The AWGN channel block adds white length Gaussian noise to the signal in this case and the General Block Deinterleaver block rearranges the bits of its input where ge is the message length and fe is the length of the vector without repeating or omitting any bits; that is, derived encoded data bits and “a” denotes the check bits a a such that ge = 2 − a − 1, fe = 2 − 1, a 3. Then the Output(bits(g)) = Input(g) input SMB is modulated using the binary phase shift keying (BPSK) method, which is a technique for The General Block Interleaver block followed by the modulating a binary signal onto a complex waveform by General Block Deinterleaver block leaves data shifting the phase of the complex signal. In digital unchanged. The LDPC decoder block is designed to baseband BPSK, the symbols 0 and 1 are modulated to decode generic binary LDPC codes. The input is a the complex numbers exp(t) and -exp(t), respectively, frame-based f x1 vector and a real-valued signal of type where t is a fixed angle. Thus for t = 0, these numbers are double. Each bit is the log-likelihood ratio for a received just 1 and -1. bit (more likely to be 0 if the log-likelihood ratio is positive). The first g bits correspond to the information The AWGN channel add white Gaussian noise to the part of a data bits. The ratio of the output sample time to input codes and it is more robust than the binary the input sample time is f/g if only the information part is symmetric channel in some specific applications because decoded, and 1 if the entire data bits is decoded. it accepts both real or complex codes and it supports multichannel input and output codes inputs as well as The BCH Decoder block recovers the message bits the frame-based processing. The Hamming decoder parses binary BCH data bits vector. The input is a frame-based through the n-bit binary position on the received SMB column vector with an integer multiple of f - bits. Each bits and decodes the data after it is sent through the group of f bits represents one data bits to be decoded. For channel. It verifies if an error is created in the original a given data bits length f, only specific message lengths g data bits by the noise in the channel, identifies the error are valid for a BCH code and f must be of the form 2M-1, and decodes the data received to the original data bits where 3< M < 16. correctly.

If f is less than 2M-1, the block assumes that the code has The system repeats this process till all n-bit binary been shortened by length 2M-1- N. However, if is greater positions is parsed and both transmitted and received than or equal to 2M-1, primitive polynomial must be codes are equivalent. After each AN-VE Parsing, the bits specified to appropriately set the value of M. The packet error rate (err) of the created SMB and received SMB is error rate is derived by computing the ratio of the total computed at z, which detects and computes the error rate number of packets received and the number of packets in of the channel using the values of the transmitted signal which the receiver detected an error. The Error Rate Tx and the received signal Rx ports. The block compares Calculation block denoted as z-1, compares input data the two signals and checks for errors. The output depicted from a transmitted signal port Tx with the input data from as sk is a vector with three entries: bit error rate, number the received signal Rx. It calculates the error rate as a of errors and total number of bits transmitted. running statistic, by dividing the total number of unequal pairs of data elements by the total number of input data elements from one source and output the error statistics.

123

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 6: Simulink of the AN-VE Communication Components.

6. RESULTS AND FINDINGS However, if more than one errors occur, the Hamming decoder block decodes incorrectly and also relatively The following table 1 and figures 7, 8and 9 show the inefficiently when sending small amounts of data, and readings of the Hamming codes, AN-VE and LDPC. The they get progressively inaccurate as the number of bits power of LDPC codes can be readily observed using increases. Significantly, for the same transmitted bits for different settings for QPSK, a default setting for rate ½, both the Hamming codes and AN-VE, it is evident that setting Es/No = 2 dB, and employing 100 decoding AN-VE performed better with lesser residual errors left iterations. With these settings it is observed that the error over in the received bits after the application of both corrective measure of LDPC fluctuates till it becomes techniques. stable. For the Hamming Coding technique, as the transmitted bits increases, the number of errors still detected after its corrective measure have been effected fluctuates.

124

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Table 1: Computational Readings Hamming Coding Technique AN-VE LDPC Transmitted Bits Number Transmitted Bits error Number Transmitted Bits Number of QPSK bits range error rate of bits range rate of bits range error errors (err) errors (err) errors rate (err) 1.013e+004 0.01076 109 1.013e+004 0.001382 14 1.62e+005 0.4915 7.963e+004 ¼ 1.018e+004 0.01041 106 1.018e+004 0.001081 11 2.16a+005 0.4928 1.064e+005 1/3 1.02e+004 0.0101 103 1.02e+004 0.00103 2 2.592e+005 0.4946 1.282e+005 2/5 1.664e+004 0.006068 101 1.664e+004 0.000201 4 4.32e+005 0.08087 3.494e+004 2/3 2.415e+004 0.00414 100 2.415e+004 0.000207 5 4.86e+005 0.101 4.909e+004 ¾ 5.007e+004 0.001997 100 5.007e+004 0.0002397 12 5.184e+005 0.1034 5.362e+004 4/5 8151 0.01227 100 8151 0.00114 9 5.4e+005 0.104 5.615e+004 5/6 9.474e+004 0.001066 101 9.474e+004 0.0057 54 5.76e+005 0.1038 5.978e+004 8/9 9036 0.01162 105 9036 0.001771 16 5.832e+005 0.1038 6.05e+004 9/10

Figure 7: Readings of Error Data (Hamming Codes) after Computation Figure 9: Readings of Error Data (LDPC Codes) after Computation

7. CONCLUSION AN-VE Codes Each error detecting and corrective technique have its own tradeoffs in terms of strength in detecting errors, 60 potency in correcting errors, portability, and ease of use. We proposed a detecting and correcting technique that is 50 more robust because although Hamming coding 40 technique is quite useful in cases where only a single error is of significant probability, the technique can only 30 be used to detect up to two simultaneous bit errors and 20 correct single errors, both cannot be done simultaneously. Also because the technique can only correct one error in Number of Error 10 each transmitted SMB, if more than one error occurs, the 0 Hamming decoder does carry the hazard of miscorrecting double errors. Conversely, AN-VE enables the detection 1.01E+04 1.02E+04 1.02E+04 1.66E+04 2.42E+04 5.01E+04 8151 9.47E+04 9036 of any number of simultaneous bit errors and corrects all errors and both can be done simultaneously. Transmitted bits Figure 8: Readings of Error Data (AN-VE Codes) after Computation

125

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Furthermore, unlike the Hamming coding technique which deals with error detection after data transmission, Partridge C., Hughes J., and Stone J. (1995). AN-VE addresses error detection right from the “Performance of checksums and CRCs over transmitter domain. AN-VE can detect the erroneous bit real data,” ACM SIGCOMM Computer in a transmitted code because every transmitted code is Communication Review. Proceedings of the repeated several times in order to verify its accuracy. Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication, vol. 25, no. 4, pp. 68–76. REFERENCES Plummer W. W. (1989). “TCP checksum function design,” ACM SIGCOMM Computer Baicheva T., Dodunekov S. and Kazakov P. (2000). On Communication Review, vol. 19, no. 2, pp. 95– the Undetected Error Probability Performance 101, Apr. 1989. of Cyclic Redundancy-Check Codes of 16-bit Reed-Solomon (2011). Reed-Solomon error correction. Redundancy, IEE Proc. Communications, vol. Available at: 147, No 5, pp. 253-256. http://en.wikipedia.org/wiki/Reed- Baicheva T., Dodunekov S., and Kazakov P. (1998). “On Solomon_error _correction.. the cyclic redundancy-check codes with 8-bit Richardson,T. J., Shokrollahi, M. A., and Urbanke R. redundancy,” Computer Communications, vol. L.(2001). “Design of capacity approaching 21, pp. 1030–1033. irregular low-density parity-check codes," Berrou, C ., Glavieux A and Thitimajshima P. (1993). ” IEEE Trans. on Information Theory, vol. 47, Near Shannon limit error-correcting coding and pp. 619-637. decoding: Turbo-codes " Proceedings of the Sheinwald D., Satran J., Thaler P., and Cavanna V. 1993 International Conference on (2002). “Internet protocol small computer Communications, pp. 1064-1069. system interface (iSCSI) cyclic redundancy Clarke, K.P. (2002). "Reed-Solomon error correction", check (CRC) / checksum considerations,” Research & development British Broadcasting Network Working Group Request for Corporation, WHP 031 BBC. Comments (RFC) 3385. Egwali A. O. and Akwukwuma V. V. N. (2011). “AN-VE: Stone J. and Partridge C. (2000). “When the CRC and An Improved Hamming Coding Technique”. TCP checksum disagree,” ACM SIGCOMM Proceedings of the International Conference on Computer Communication Review. ICT for Africa 2011. March 23 - 26. Vol. 3. Proceedings of the Conference on Applications, Pg. 9 - 16. Covenant University and the Bells Technologies, Architectures, and Protocols for University of Technology. Ota, Ogun State, Computer Communication, vol. 30, no. 4, pp. Nigeria. 309–319. Feldmeier D. C. (1995). “Fast Software Implementation Stone J., Greenwald M., Partridge C., and Hughes J. of Error Detection Codes,” IEEE/ACM Trans. (1998). “Performance of Checksums and Networking, vol. 3, no. 6, pp. 640-651. CRC’s over Real Data,” IEEE/ACM Trans. Forney, G. D., (1966).”Performance of concatenated Networking, vol. 6, no. 5, pp. 529-543. codes," Concatenated Codes, pp.16, 8290S. Kazakov P. (2001). “Fast calculation of the number of minimum-weight words of crc codes,” IEEE Transactions on Information Theory, vol. 47, no. 3, pp. 1190–1195. Koopman P. (2002). “32-bit cyclic redundancy codes for internet applications,” in International Conference on Dependable Systems and Networks 2002, June 23–26, pp. 459–468. Koopman P.and Chakravarty T. (2004). “Cyclic redundancy code (CRC) polynomial selection for embedded networks,” in International Conference on Dependable Systems and Networks 2004, pp. 145–154. McAuley A. J. (1994). “Weighted sum codes for error detection and their comparison with existing codes,” IEEE/ACM Trans. on Networking, vol. 2, no. 1, pp. 16–22.

126

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Evolutionary Continuous Genetic Algorithm for Clinical Decision Support System

Prem Pal Singh Tomar & Ranjit Singh1, Faculty of Engineering Dayalbagh Educational Institute Agra, India [email protected], [email protected]

ABSTRACT The Medical Multimedia based Clinical Decision Support System (MM-CDSS) using the continuous genetic algorithm (CGA) methodology presents a foundation for a new technology of building intelligent computer aided diagnosis systems. The system has trained set of 280 cases for Indian heart patients for four major heart diseases: coronary heart disease, rheumatic valvular heart disease, chronic cor pulmonale, and congenital heart disease, which contain 24 symptoms for each heart disease. Using the CGA methodology, 24 critical symptom values have been identified. The best chromosome is obtained through matlabR2007a simulation. This paper presents a evolutionary continuous genetic algorithm for MM-CDSS, supporting diagnosis of four major heart diseases from their symptoms and signs through employing Microsoft Visual Basic .NET 2005 along with Microsoft SQL server 2005 environment with the advantage of Object Oriented Programming technology. A Consultant physician’s interpretation was used to evaluate the system’s Sensitivity, Specificity, Positive Prediction Value and the Negative Prediction Value. The preliminary results showed promising usage for the MM-CDSS in terms of correct and accurate diagnosis for the inexperienced physician as well as consistent and timely diagnoses, in the study of diagnostic protocol, education, self-assessment, and quality control of four major heart diseases that were investigated.

Key words: Medical Multimedia based Clinical Decision Support System, Heart Diseases Diagnosis, Continuous Genetic Algorithm, Diagnostic Features, Physician. African Journal of Computing & ICT Reference Format Prem Pal Singh Tomar1 , Ranjit Singh1, (2013). Evolutionary Continuous Genetic Algorithm for Clinical Decision Support System. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 127-140

1. INTRODUCTION For the last few decades, significant efforts have been In clinical practice, making decision involves a careful made in the field of research dedicated to using different analysis of harms and benefits associated with different data mining and machine learning techniques to discover treatment options. These decisions, often associated with useful medical knowledge and rules [1-4]. In the different high stake and important long term consequences, are data mining and machine learning techniques, the genetic frequently made in presence of limited resources and algorithms (GAs) have been accepted to be dominant in information and an incomplete clinical picture. Under medical knowledge discovery. Genetic algorithms are such circumstances, a rigorous and objective analysis of basically a search algorithms developed, following the outcomes and probabilities is essential to achieve the best principles of natural selection and natural genetics [5, 6], possible decision given a specific clinical situation. and have been successful in complex knowledge Therefore, physician is required to be fully conversant discovery and rule extraction problems. For example, with the diversity of possible patterns, recognize and diagnosis of hypertension based on the applied genetic diagnose them, timely and accurately. Hence, a physician algorithm was presented by Adel [7] using geometrical who is not a specialist in the pathology of the heart parameters. diseases has to refer to textbooks and study past diagnosis before concrete diagnosis can be made and conclusion Anbarasi, Anupriya and Iyengar [8] proposed a procedure reached. Hence, there is the need for a system, which can using genetic algorithm and Decision Tree data mining assist the physician to reach timely and accurate decision. technique to reduce the number of tests which were needed to be taken by a heart patient. Kiran and Ramesh [9] proposed a framework supported with genetic evolution to predict the heart attack. Kurzynski and Zolnierek [10] developed and evaluated methods for performing Sequential Classification (SC) using fuzzy rules and an optimization procedure using the

127

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Real-coded Genetic Algorithm applied to the computer- The pseudo code for Continuous Genetic Algorithm of aided recognition of patient's acid-base equilibrium GA platform is given as under: states. Begin (1) In this paper Section 2.0 briefly describes the methodology of proposed continuous genetic algorithm Generate initial population Npop with R random for critical diagnostic symptoms value. Symptom chromosomes with m x n real number (range -1 to 1) acquisition from the system user is provided in Section matrix. 3.0 Finally, Section 4.0 presents the Results and Discussion. Evaluate Fitness of all chromosomes in population Npop

While stopping condition is not true do 2. CGA METHODOLOGY DESCRIPTIONS Begin (2) Continuous genetic algorithm which is a computerized stochastic search and optimization method that works by Rank the chromosomes in Npop in mimicking the evolutionary principles and chromosomal ascending order of their fitness processing in natural genetics. Solutions from a population are used to form a new population. Solutions Keep the best R/2 chromosomes in that will form new solutions are selected according to Npop as elite population their fitness: the more suitable they are, the more chances they have to reproduce. For R/2 chromosomes

When the variables are continuous, it is more logical to Begin (3) represent them by floating-point numbers. In addition, since the binary GA has its precision limited by the Select two parent binary representation of variables, using floating point chromosomes P1 and P2 numbers instead easily allows representation to the with roulette wheel strategy machine precision. This continuous GA (CGA) also has the advantage of requiring less storage than the binary Crossover operation (P1, GA because a single floating-point number represents the P2) variable instead of Nbits integers. The continuous GA is Mutation operation inherently faster than the binary GA, because the chromosomes do not have to be decoded prior to the End (3) evaluation of the cost function. Evaluate offspring population Then three standard genetic operations, i.e., selection, crossover, and mutation are performed to produce a new Replace Npop with elite population + generation. Such procedures are repeated until the pre- offspring population specified number of generations is achieved, or the required accuracy is satisfied.The Continuous GA If terminate condition is True then algorithm for selecting critical features values for Escape diseases diagnosis is given below. The algorithm will be stopped when the number of iterations is exceeded. It will End (2) also terminate when a chromosome in the population can recognize all the instances in the data set successfully. Output and save the best chromosome as solution in the When the algorithm is stopped, it will output the best database chromosome in Npop. End (1)

128

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2.1 Encoding of Diagnostic Features Table 1: Parameters used in the proposed continuous Each heart disease case includes 24 important recorded genetic algorithm diagnostic features as shown in table-1, which are believed as critical information and symptoms required 2.3 Initial Population Generation for diagnosing the chosen 4 chronic heart diseases: Generate initial population Npop with R random coronary heart disease, rheumatic valvular heart disease, chromosomes with m x n continuous value number chronic cor pulmonale, and congenital heart disease. (range -1 to 1) matrix, where m and n represents the Diagnostic symptoms /facts and Chronic Heart or Lung The diagnostic features are encoded using the three-value Diseases respectively. A chromosome C of R randomly ordinal scales in terms of the degree of the seriousness, generated chromosomes, having gene values varying in The encoded value ‘0’ represents the absence of the the range of (-1, 1) randomly as shown below. attribute, ‘0.5’ represents intermediate level of the attribute and ‘1’ represents the largest presence of the attribute as shown in table 2. C = , g ϵ [-1, 1] , ij 2.2 Setting of the Parameters The Continuous GA algorithm (CGA) for selecting critical features values for 4 chronic heart diseases 1 ≤ i ≤ m and 1 ≤ j ≤ n diagnosis is given below. The algorithm will be stopped when the number of iterations is exceeded. It will also Where, gene gij represents the ith Diagnostic symptom of terminate when a chromosome in the population can the jth Chronic Heart or Lung Disease. The value and the recognize all the instances in the data set successfully. sign of a gene indicate the relationship between the When the algorithm is stopped, it will output the best Diagnostic Symptom and its corresponding Chronic Heart Diseases. chromosome in Npop. The parameters used in the proposed CGA algorithm are listed in Table 1 There are 4 diseases decision targets and 24 numbers of diagnostic features for each disease. A randomly Parameters Value generated chromosome C, having gene values varying in Initial Population size 64 the range of (-1, 1) randomly, chromosome C is then Elite population selection rate (Xrate) 50% represented as [24 x 4] real number matrix as shown in Number of generations 100 Figure 1, the matlabR2007a command window. Crossover probability, pc 0.5 Mutation probability, pm 2%

Figure 1: A Chromosome ‘C’ in CGA Model

129

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

First, second, third and forth column in figure above It will also terminate when a chromosome in the represents the chronic cor pulmonale, congenital heart population can recognize all the instances in the data set disease, rheumatic valvular heart disease, and coronary successfully. If the termination condition is not satisfied, heart disease respectively. The value and the sign of a then select and keep the best (Z=R/2) chromosomes in gene indicate the relationship between the diagnostic initial population Npop as elite population (The feature and its corresponding chronic heart disease. chromosome with the best fitness is kept from generation to generation) i.e. only the best chromosomes are selected 2.3 Chromosome Fitness Evaluation to continue, while rests of the chromosomes are deleted, Evaluate Fitness (A value associated with a chromosome Where 1 ≤ Z ≤ R/2. The selection rate, denoted by Xrate, that assigns a relative merit to that chromosome) of all is the fraction of Npop that survives for the next step of the chromosomes in initial population Npop. The fitness mating. function evaluates the performance of each chromosome to measure how close it is to the solution. A fitness 2.4 Genetic operations function in CGAs is a particular type of objective Genetic operators used in genetic algorithms maintain function that quantifies the optimality (i.e. extent of genetic diversity. Genetic diversity or variation is a “fitness” to the objective function) of a chromosome necessity for the process of evolution. Genetic operators (solution) so that that particular chromosome may be are analogous to those which occur in the natural world: ranked according to its fitness value against all the other reproduction (i.e. selection), crossover (i.e. chromosomes. In our diagnostic problem, it corresponds recombination), and mutation. to the number of correct classifications over the whole dataset. Denote an instance At (1 ≤ t ≤ X) in the data set 2.4.1 Selection of Parent Chromosomes with X instances as (at1, at2,. . ., atk,. . ., atm), where atk (1 Select two parent chromosomes P1 and P2 with Roulette ≤ k ≤ m) is the value of Diagnostic Symptoms / facts k in wheel strategy. Two chromosomes are selected from the instance At. mating pool of elite chromosomes to produce two new offspring chromosomes. Pairing takes place in the mating population until elite population offspring are born to replace the discarded chromosomes.

Assume the population Npop has R chromosomes, for each chromosome Cy, where 1 ≤ y ≤ R. The average The classification result λRt of a chromosome C to fitness function f(Cav) is given by instance At can be represented as:

………(3) ………(1)

The expected count (ec) of a chromosome y is given by If the classification result λRt is the same as the ground truth ζTt of the instance At, the indication function for the instance, denoted as ξt, has a value of 1; otherwise, the value is 0. Thus, for chromosome C, its fitness value is …………..(4) obtained as: The probability of selection, γs(Cy), is computed as:

………(2) ………….(5) Where, f(C) = Fitness value of chromosome C In Roulette wheel selection, a chromosome Cy is selected ξt = indication function for the instance if a uniformly generated random number µ in [0, 1] X= instances in the data set satisfies the following equation:

Rank the chromosomes in initial population Npop in ………(6) ascending order of their fitness so that the elite chromosomes can be separated easily from the initial where =0 for k=0 population N Check if the termination conditions are pop. satisfied. The algorithm will be stopped when the number The steps of Selection operation are as under: of iterations is exceeded.

130

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

1. Calculate the Average Fitness Value from the fitness function 2. Calculate the Expected Count = Fitness Function/ Average Fitness Value 3. Probability of Selection (γs) = Expected count / No. of chromosomes 4. Calculate the cumulative probability (cpx)= γsx+ cpx-1 5. Chromosome selection is done by seeing the random number µ in cumulative probability (cp)

Figure 2: Selection Operation in CGA Model

As shown in the Figure 2, the first column denotes the chromosome identification number and its fitness value in second column according to the elite chromosome table, third, forth, fifth, sixth and seventh column denotes the expected count value, probability of selection, cumulative probability, random generated number and selected chromosome for mating process of continuous genetic algorithm.

2.4.2 Crossover of Parent Chromosomes Two parents Figure 3 and Figure 4, mate to produce two offspring. The basic operator for producing new chromosomes in the CGA is that of crossover (recombination). Similarly as its counterpart in nature, crossover operation produces new individuals that have some parts of both parent’s genetic material. The parents have produced a total of elite population offspring, so the total chromosome population is now back to Npop.

If two parent chromosomes P1 and P2 are selected, they perform crossover operation with a crossover probability (cp) to generate two new chromosomes ch1 and ch2 through the following way: ch1= S.P1 + (1-S).P2 ch2= (1-S).P1 + S.P2

131

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Where S is a Stochastic matrix (Figure 5) with each element valued randomly between [0, 1] and the operator “.” denotes the element-by-element matrix multiplication.

Figure 3: Parent chromosome P1 in GA Model

Figure 4: Parent chromosome P2 in GA Model

132

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 5: Stochastic matrix ‘S’ in GA Model

An example of the crossover operation is illustrated below:

If two parent chromosomes P1 and P2 are selected, they perform crossover operation with a crossover probability pc to generate two new child chromosomes ch1 (Figure 6) and ch2 (Figure 7) by using the crossover operation.

Figure 6: child chromosome ch1 in GA Model

133

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 7: Child Chromosome Ch2 in GA Model

2.4.3 Mutation Operation Mutation is a random process where one allele of a gene is replaced by another to produce a new genetic structure. In CGAs, mutation is randomly applied with low probability value, and modifies elements in the chromosomes. To perform the mutation operation the random substitution method is adopted i.e. the chromosome chosen to mutate is replaced by a new randomly generated chromosome C (Figure 8), having gene values varying in the range of (-1, 1) of matrix size [m x n].

Figure 8: Randomly Generated Chromosome for Random Substitution in GA Model

134

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2.4.4 Offspring Fitness Evaluation A fitness function evaluates the fitness value of each child chromosome to measure how close it is to the solution in the similar manner as explained. To maintain the size of the constant original population, the numbers of new individuals are added to the numbers of elite chromosomes.

2.4.5 Best Chromosome Selection Continue the iteration till the average fitness of chromosomes increases. Having the maximum value of fitness function, the resulting chromosome will be the answer as best chromosome. As the best chromosome is the best candidate solution, and a gene is numerically encoded, the value of each gene indicates the relationship between each diagnostic feature and its corresponding goal disease. A positive gene indicates that the corresponding feature supports the diagnosis of the disease i.e. higher value indicate the higher effect of that feature on the diagnosis; or a negative gene indicates no effect of the feature for diagnosis. If the value of a gene is negative, it is set to “0” for convenience, as the feature itself in this case has no meaningful implication in the disease classification. The evolved best chromosome (Figure 9) by the proposed continuous genetic algorithm is saved to the database.

Figure 9: Best Chromosome

3. SYMPTOM ACQUISITION FROM THE SYSTEM USER

The details about the diagnostic symptoms values (either The Diagnostic symptoms values are multiplied with 1, 0.5 or 0) are collected from the system users using a each column of the best chromosome and then add the specially graphically designed interactive interface in multiplied values to obtain the final diagnostic resultant Microsoft Visual Basic .Net 2005 platform as shown in value corresponding to a chronic heart disease that is table 3 [Figure 2(1) – Figure 2(9)], for the design and displayed to the system user through a graphical user development of a CGA Module with the advantage of interface of CGA platform as shown in Figure 10. Object Oriented Programming technology. The Microsoft SQL server 2005 is used to develop the database for different 4 chronic heart diseases.

135

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4. RESULTS AND DISCUSSION REFERENCES

In this research work 280 diagnosed Chronic Heart [1] Vera, M., A. Bravo and R. Medina, (2010), disease cases suffering from Coronary Artery Heart “Myocardial border detection from Disease, Rheumatic Valvular Heart Disease, Chronic Cor ventriculograms using support vector machines Pulmonale Heart Disease and Congenital Heart Disease and real-coded genetic algorithms”, Comput were acquired from specialized database available at Bio Biol Med., 40(4), 446-455. Medical Engineering Research (BMER) Lab of Faculty [2] Mahanand, B. S., S. Suresh, N. Sundararajan and K. of Engineering, D.E.I., Dayalbagh, Agra, India, of M. Aswatha, (2012), “Identification of brain duration April 2004 to May 2011 and from Dr. Varun regions responsible for Alzheimer's disease Chaudhary (Physician), M.B.B.S., M.D., Heart, Chest using a Self-adaptive Resource Allocation and Allergy Research Laboratory, Agra, India. Network”, Neural Netw., 32, 313-22.

[3] Park, Y.J., S. H. Chun and B. C. Kim, (2011), “Cost- The 273 subjects suffering from Chronic Heart disease sensitive case-based reasoning using a genetic were selected for medical investigations and diagnosis at algorithm: Application to medical diagnosis”, Bio Medical Engineering Research (BMER) Lab of Artif Intell Med., 51(2), 133-145. Faculty of Engineering, D.E.I., Dayalbagh, Agra, India [4] Raghuwanshi, M.M. and O.G. Kakde,(2007), and Heart, Chest and Allergy Research Laboratory, New “Distributed quasi steady-state genetic Agra, Dayalbagh Road, Agra, India. algorithm with niches and species”,

International Journal of Computational Table 4 shows the diagnostic result for True Positives Intelligence Research, 3(2), 155-164. (TP), True Negatives (TN), False Positives (FP), and [5] Goldberg, D.E., Genetic Algorithms in Search, False Negatives (FN) from the developed CGA Model. Optimization and Machine Learning, Addison-

Wesley, Reading MA, 1989. The following equations are used to calculate Sensitivity [6] Davis, L., Handbook of Genetic Algorithms, Van (S ) and Specificity (S ) [11] of CGA Model (Table 5): e p Nostrand Reinhold, Prentice Hall. New York,

1991. [7] Adel, A. S., (2007), “Diagnosis of Hypertension based on the Applied Genetic Algorithm of Retinal Vascular Tree”, ICGST- BIME Journal, 7(1), 123-129.

[8] Anbarasi, M., E. Anupriya and N.C.H.S.N. Iyengar, (2010), “Enhanced Prediction of Heart Disease with Feature Subset Selection using Genetic Algorithm,” International Journal of

Engineering Science and Technology, 2(10), 5370-5376. [9] Kiran, P. Sree and I. Ramesh Babu, (2011), “Non Linear Cellular Automata in Predicting Heart Attack”, International Journal of Hybrid Information Technology, 4(1), 73-79. [10] Kurzynski, M. and A. Zolnierek, (2006), “Sequential classification via fuzzy relations”, Artificial Intelligence and Soft Computing, Springer, 4029, 623–632. [11] Tomar, P.P.S. , R. Singh, and P. K. Saxena, “Multimedia Based MDSS For Chronic Lung Diseases Diagnosis Using Rule Based Technique”, Afr. J. of Comp & ICT., IEEE Nigeria, 4, 3 (2), 2011, 1-6.

136

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

APPENDICES

The attributes and their encoding values of chronic heart disease diagnostic symptoms

Serial No. Chief Symptoms Attributes and their corresponding encoding values

1 0.5 0 1. Edema Serious Medium No 2. Breathlessness Serious Medium No 3. Fatigue Serious Medium No 4. Angina Serious Medium No 5. Heart murmur Serious Medium No 6. Syncope Serious Medium No 7. Chronic cough(COPD) Serious Medium No 8. Hepatomegaly Serious Medium No 9. Hypoxia Serious Medium No 10. Pulmonary hypertension Serious Medium No 11. Hypertrophy & dilation of RV Serious Medium No 12. Cyanosis Serious Medium No 13. Growth retardation Serious Medium No 14. Fever Serious Medium No 15. Sore throat Serious Medium No 16. Arthalgia Serious Medium No 17. Nausea Serious Medium No 18. Pericardial rub Serious Medium No 19. Anxiety Serious Medium No 20. Family history Serious Medium No 21. High BP Serious Medium No 22. Diabetes (blood suger) Serious Medium No 23. Smoking Serious Medium No 24. High cholesterol Serious Medium No

Table 3: Symptom acquisition from the system user

Figure (1): Interactive Interface for acquisition of Edema Figure (2): Interactive Interface for acquisition of Diagnostic Symptom in CGA Model Breathlessness Diagnostic Symptom in CGA Model

137

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure (3): Interactive Interface for acquisition of Fatique Figure (4): Interactive Interface for acquisition of Chest Diagnostic Symptom in CGA Model Discomfort Diagnostic Symptom in CGA Model

Figure (5): Interactive Interface for acquisition of Chronic Figure (6): Interactive Interface for acquisition of Cough Diagnostic Symptom in CGA Model Hepatomagaly Diagnostic Symptom in CGA Model

Figure (7): Interactive Interface for acquisition of Hypoxia Figure (8): Interactive Interface for acquisition of Pulmonary Diagnostic Symptom in CGA Model Hypertension Diagnostic in CGA Model

138

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure (9): Interactive Interface for acquisition of Dilation od Right Ventricular Diagnostic Symptom in CGA Model

Figure 10: Diagnostic Results for Heart Diseases using CGA Model with Multimedia support tools

139

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Table 4: Diagnostic Result from CGA Model

Heart Total

Diseases

coronary heart heart coronary disease rheumatic heart valvular disease cor chronic pulmonale heart congenital disease Coronary Heart Disease 76 2 3 1 82 Rheumatic Valvular Heart Disease 1 53 3 2 59 Chronic Cor Pulmonale 0 2 65 3 70 Congenital Heart Disease 1 3 3 55 62 Total 273

Table 5: Sensitivity (Se) and Specificity (Sp) Results for CGA Model

Heart CGA Model Diseases

Se % Sp % Coronary 92.68 98.95 Heart Disease Rheumatic Valvular Heart Disease 89.83 96.72

Chronic Cor Pulmonale 92.85 95.56 Heart Disease Congenital 88.71 97.15 Heart Disease Average 91.01 97.09

140

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Mass Connectivity based Digital Nervous System for Nigeria

Monica N. Agu Department of Computer Science University of Nigeria Nsukka Nigeria. Email: [email protected] +2348039329480

ABSTRACT Experience has shown there are more than enough data with little or no information avaliable. Nigeria being a country endowed with a lot of resources and whose people is about 170,123,740 million[6] needs a way of getting its populace with the necessary information they need. This can only be achieved through connecting people to all their information needs. Despite the huge investment and advancement in ICT, a lot of data has been generated but can be stored, processed and disseminated. With ICT , Nigeria will evolve to the next level of mass connectivity which is the digital nervous system. In this ICT is used to connect all information to the users. This looks at having a repository of information which is diseminated to people through electronic village halls with trained social workers, gsm operators who broadcast certain information to all citizens. The people’s responses are sent to digital nervous system through the social workers.

Keywords: Digital Nervous System, ICT, Mass connectivity . African Journal of Computing & ICT Reference Format Monica N. Agu (2013). Mass Connectivity based Digital Nervous System for Nigeria. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 141-146

1. INTRODUCTION The question then arises , what is digital nervous system. Steve Ballmer in his attempt to explain it likened it to Vast quantities of information exist in almost every human nervous system which allows us to hear, see and sector in this country, Nigeria but are scattered about in take input. This also allows us to think, plan and make different locations and in different formats for use by decisions and communicate and take action[3]. In this, different groups of people for different purposes. This just like the biological nervous system we have the format has hindered the citizens especially the rural information we need. We are always alert to the most dwellers from getting the valuable information to lift important and block out information that are not them up from the cages where they find themseleves. important ie information that are valuable getting to These rural dwellers have no confidence in themselves. people who need to know about them[4]. In another work There is no way of interacting with those outside their it is described as being synonymous with the term zero group thereby increasing the digital divide. Thier latency enterprise ie the way an enterprise use IT system potentials can not be identified because there is no to rapidly communicate between customers, employers information about them. There is need to have a reform and trading partners[5]. and also to make a more efficient use of the country’s valuable information. Gates in his book 'Business @ The Speed Of Thought’ defined Digital nervous system as “Futuristic vision of a 2. RELATED LITERATURE. computer network that imitates the biological nervous system in (1) sifting what bit of incoming information is In Nigeria we have more than enough data but despite important from what is not, (2) learning from experience, this no information is avaliable for the masses that need (3) adapting to changes in its external environment, and them. This is because the huge amount of data generated (4) reacting swiftly to advantageous or threatening are not properly stored, processed and diseminated situations in addition to managing the internal despite the huge advancement in ICT. It is strongly environment of an organization. Gates further used believed that with ICT, the Nigerian populace especially Digital nervous system to describe a vision for how the the rural dewellers will evolve to the next level of IT infrastructure of an enterprise could be analogous to information age which is the digital nervous system. the autonomic nervous system of a biological organism[1].

141

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The nature of digital nervous system was illustrated by The data from this module is sent to the central module Merrill Lynch which was brought on line in 1998 in from where the information needed is sent to the which their financial consultants(FC) can automatically populace through their village halls by their display what ever real-time data in which they call upon intermediaries such as social workers who have direct to monitor on regular basis. The FC’s were given several access to the central module. The poliferation of the use views of the relevant data and they can at a glance vary of GSM phones even in the hinter lands makes the key decision variables. concept easily achievable. Any feedback from the citezenery is also sent back to the source. The same unsatisfactory conditions which existed at Merrill Lynch in1997 exist almost every where in Nigeria Health Education: This module contains data about today and this can be corrected in a similar way. Vast health. Health education is educating people about quantities of information exist in the bowels of every health[8] which contains many areas and they include state ministries already in electronic forms but are environmental health, physical health, social health, scattered in different locations for use by different local emotional health, intellectual health, and spiritual government areas by different people for different health.[9] It can be defined as the principle by which purposes. The data typically reside in state ministries- individuals and groups of people learn to behave in a education, health, etc. There is therefore need to construct manner conducive to the promotion, maintenance, or a digital nervous system capable of accessing all such restoration of health. Since we can define Health in many data from a single point and extracting from it the needed ways , so also we can define health education in many information for the Nigerian populace. ways. This has led The Joint Committee on Health Education and Promotion Terminology of 2001 to define Health Education as "any combination of planned 3. DESIGN METHODOLOGY learning experiences based on sound theories that provide

individuals, groups, and communities the opportunity to In designing this, there will be a central module where acquire information and the skills needed to make quality all the other modules are attached which is the repository health decisions."[10] of information from where information is retrieved and response sent back to the central module. This can be shown using fig 1. In fig 1 there are six modules The World Health Organization defined Health connected to the central module and they include the Education as "comprising of consciously constructed opportunities for learning involving some form of following : communication designed to improve health literacy, 1. Talent Hunt including improving knowledge, and developing life 2. Health Education skills which are conducive to individual and community 3. Poverty Reduction Initiatives health." [11] The purpose of this module is to positively 4. Fighting Pandemics influence the health behavior of individuals and 5. Empowerment Initiatives 6. Political Enlightenment communities as well as the living and working conditions 7. that influence their health. The information got from this module will help in improving the health status of Talent Hunt : This is “Getting the People You Need, individuals, families, communities, states, and the nation. When You Need them”. One of the biggest challenge of It will also enhance the quality of life for all people, any cheif executive is always finding and keeping good reduce premature deaths. people. Many people are talented but are disadvantaged to the extent that they never put their talent to use. With The costs (both financial and human) that individuals, this connectivity there could be a continous search for employers, families, insurance companies, medical talented but disadvantaged people who could be helped to develop their respective talents to the advantage of facilities, communities, the state and the nation would themselves, their families, their LGA’s, states and the spend on medical treatment can be reduce when data in this module focuses on prevention. This can only be country as a whole. This module deals with every achieved by the use of IT by connecting people with the information about ministries that require people with the required information. The data from this module is kind of qualifications needed for that particular position. connected to the central module which is the repository

of all information. IT will help the health workers reduce

the amount of paper work done, and improve their data

accuracy. It will also empower these health workers to provide timely care and information. These health workers will educate people on family welfare etc.

142

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Fig. 1: The Digital Nervous System for Empowering the Nigerian Populace through ICT

Poverty Reduction Initiatives: This module also Programme (PAP), Better life Programme (BLP) contains all sorts of data concerning poverty reduction etc.[14]. These initiatives failed because the constant initiatives. This includes the summary of suggestions needed pieces of information were not avaliable and which several Government and non-Government have where avaliable were not accessible by the people who given poverty reduction initiatives. In Nigeria a lot of needed them. For any of these intiatives to achieve its poverty reduction intiatives have been used. Some objective, there must be a flow of information needed to examples are People’s Bank of Nigeria(PBN), Family sustain the programme from the government and NGO’s support Programme(FSP), Poverty Alleviation to the people , infact vice versa.

143

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

With this digital connectivity the information can be Empowerment Initiatives: The module contains IT- online and displayed in the Electronic village halls based empowerment initiatives which includes on-line thereby making such information accessible to the rural jobs one can do in virtual offices. This also involves the dwellers. These information will be better understood if provision of IT equipment or tools to enhance the flow of there is an intermediary to explain in greater details what information in a sustainable way. The tools include is required of them. This can only be achieved with village pay phones which can empower the poor by trained social workers. Also GSM operators can creating job, improving access to health care, education broadcast certain information to all, and the citizens and other services[12]. This village pay phones provide a responses are directed to the DNS or via the social means of communication and income generation. Also workers. another tool is Telecentres which provide access to information on community level. Many of these Fighting Pandermics: This module contains data telecentres combine radio, phone,fax, email and internet concerning different types of dieseases their symsptons, facilities in rural communities. The telecentres provide and the medications needed for each. This can be information on things like agriculture, health education achieved through telecentres, telephones, radio, and livelyhood. In [13] the author gives example of how televisions and the internet. The poor can be informed telecentres can be a strong health information about diseases like HIV/AIDS, Bird flu; Influenza, disseminating tool and how the youth have improved Cholera and this will help them come out of poverty access to information through community based since poverty is not only lack of money but lack of the telecentres in Zambia. necessary information. This can be achieved through the radio and internet. In Sri Lanka radio is used as an interface between rural poor and the internet. In this a The Central Module: This module is the repository of panel of resource persons browse the internet in response information. Any information from the other modules are to listeners request and this is relayed to the poor through sent to the people through Electronic Village Halls with a radio program on a daily basis[15]. trained social workers, GSM operators to broadcast certain information to all. The feedback from the citizens Political Enlightenment: In this module different mass are sent directly to the DNS or through the social enlightenment campaigns are listed. This will give the workers. people every information about the politics of the country. The masses have the oppurtunity of knowing who wants to govern them and those to cast their votes for. This will expand the economy,will make politics more diverse and by default more corupt. This when it happens and people are impacted with a lot of ideas they will start discussing new ideas for society. The populace begin to finally think outside the box they had been enclosed in for so long and they now develop new opinions on everything in society and government. The "Age of Enlightenment" became much expanded not only knowledge but to greater movement of greater capital (money) and ideas. These were part of the contributing factors to the rise of the middle class and a decline in the power of the aristocrats. With this people began to question a lot of things. It's kind of a "feedback loop", where more knowledge leads to more change which in turn leads to more knowledge, etc. The Enlightenment, or Age of Enlightenment, rearranged politics and government in earthshaking ways.

144

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4. CONCLUSION. With the use of GSM, Electronic village halls and Internet connectivity a lot of things can be achieved Despite the huge investment and advancement in ICT, a namely: lot of data has been generated but can be stored, a. There will be oppurtunity creation for the processed and disseminated. With ICT , Nigerian citizens populace especially the rural dewellers will evolve to the b. There will be identification and development of next level of mass connectivity which is the digital potentials for the citizens. nervous system. In this ICT is used to connect all c. The people’s mind set will be transformed to information to the users.In Nigeria we have more than one of positive values, self confidence, honesty enough data but despite this no information is avaliable and intergrity. for the masses that need them. This is because the huge d. This will create powerful ICT based amount of data generated are not properly stored, communities to facilitate mutual beneficial processed and diseminated despite the huge advancement interaction both among groups within nations in ICT. In this DNS information which people are not and across large geographical divides. aware of are brought to them. e. Use of GSM and internet to dynamically recieve a wealth of new business ideas from the This information can be packaged for them because they populace to enable the same information to be have ICT to help them. This is achieved through equiping ready to those they can benefit. their villages/communities with televisions. When there f. There will be business and engineering skills is proper mass connectivity the society is majorly training information driven. Rapid cross pollination of ideas g. The quality of life in rural Nigeria will be becomes possible. This would lead to spontaneous enhanced through the effective use of the mass creation of new business ideas and markets which benefit connectivity thereby bridging the digital divide. many of the citizens. Also the wide spread of GSM Also when this connectivity is properly phones and the internet opens the door for a wide range harnessed, democracy dividends will be taken of people to subscribe to and access information of to the hinter lands. interest to them. h. A new employment oppurtunities that enhances income levels will be created when there is a well articulated pursuit of professional development, bussiness development, high technology apprenticeship and the repacking of the individual in a way that enhances self worth.

This DNS gives room for people to send their feedback through the social workers who have direct access to the database. The GSM operators also broadcast the necessary information to the populace. In this way nation will be connected and every information that comes from the different modules will be sent to the populace. This in effect will bridge the digital divide thereby empowering the rural poor in Nigeria. With this Nigeria will grow into a better nation with its citizens being aware of what is happening ie having the necessary information they need.

145

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFERENCES

[1] Gates, Bill (1999). Business @ The Speed of [12] Grameen Foundation. Empowering People, Thought, Warner Books, New York. ISBN 0- changing Lives, Innovating for the World’s 446-67596-2 poor, 2001. http://www.businessdictionary.com/definition/d http://www.grameenfoundation.org igital-nervous system.html#ixzz1tjauvJOG [13] Gerster,R and Zimmermann, S. Information [2] http://www.theregister.co.uk/1999/04/12/digital and Communication Technologies (ICT) for _nervous_system_not_bill/ Poverty Reduction; discussion Paper, 2003 [3] Steve Ballmer Speech Transcript - Digital Nervous System Seminar [14] CBN. Annual Report and Statement of [4] Digital Nervous System 10/98 Accounts; 2003 http://www.roughnotes.com/rnmagazine/199 [15] Dilante, w(2003) the use of ict’s for poverty 8/october98/10p34.htm reduction [5] "Moving Toward the Zero Latency Enterprise". www.unescap.org/rural/ICTEGNov/2003/srilanka Retrieved 2009-05-02. Dilanthe) [6] Nigeria Population - Demographics Source: CIA World Factbook - U http://www.indexmundi.com/nigeria/population .html [7] Digital Nervous Systems: Making Sense of Shared Information.SIAM News, volume 32, number 10 http://www.siam.org/pdf/news/802.pdf [8] McKenzie, J., Neiger, B., Thackeray, R. (2009).Health education can also be seen as preventive medicine(marcus 2012). Health Education and Health Promotion. Planning, Implementing, & Evaluating Health Promotion Programs. (pp. 3-4). 5th edition. San Francisco, CA: Pearson Education, Inc. [9] Donatelle, R. (2009). Promoting Healthy Behavior Change. Health: The basics. (pp. 4). 8th edition. San Francisco, CA: Pearson Education, Inc. [10] Joint Committee on Terminology. (2001). Report of the 2000 Joint Committee on Health Education and Promotion Terminology. American Journal of Health Education, 32(2), 89-103. [11] World Health Organization. (1998). List of Basic Terms. Health Promotion Glossary. (pp. 4). Retrieved May 1, 2009 from http://www.who.int/hpr/NPH/docs/hp_glossary _en.pdf.

146

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Performance Evaluation of Neural Network MLP and ANFIS models for Weather Forecasting Studies

Oyediran O. F. and Adeyemo A. B. Computer Science Department University of Ibadan [email protected], [email protected]

ABSTRACT Weather forecasting is the application of science and technology to predict the state of the atmosphere for a future time at a given location. It is carried out by collecting quantitative data about the current state of the atmosphere and past and/or present experiences. In this study Adaptive Neuro-Fuzzy Inference System (ANFIS) and Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN) models were used to analyze metrological data sets obtain from the metrological station. The data covers a ten year period (2002-2012) were for the monthly means of minimum and maximum temperature, rainfall, wind run, and relative humidity. The results showed that both models could be applied to weather prediction problems. The performance evaluation of the two models that was carried out showed that the ANFIS model yielded better results than the MLP ANN model with a lower prediction error.

Keywords: Weather Forecasting, Artificial Neural Networks, Neuro-Fuzzy Inference Systems . African Journal of Computing & ICT Reference Format Oyediran O. F. and Adeyemo A. B. (2013). Performance Evaluation of Neural Network MLP and ANFIS models for Weather Forecasting Studies. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 147-164

1. INTRODUCTION

Prediction is based on past and/or present experiences. It Data mining has been applied in many areas e.g. Health, plays a vital role in our world today as it is applied in Industry and Agriculture for good decision making as a predicting a lot of events including weather changes. result of past data collected. Today, weather prediction is Weather simply refer to the condition of air on earth at a made by collected quantitative data about current state of given place and time. The application of science and atmosphere process to project how the atmosphere will technology to predict the state of the atmosphere in future evolve (Wikipedia, 2008). Weather forecasting entails time for a given location is important due to its effect on predicting how the present state of the atmosphere will human activities. The chaotic nature of the atmosphere change. Present weather conditions are obtained by implies the need for massive computational power to ground observations, observations from ships and solve the equations that describe the atmospheric aircraft, radiosondes, Doppler radar, and satellites. This conditions. (Lai, 2004). This is as a result of the information is sent to meteorological centers where the incomplete understanding of atmospheric processes data are collected, analyzed, and made into a variety of which mean that forecasts become less accurate as the charts, maps, and graphs in facing the prediction of the difference in time between the present moment and the weather assumed that the state of the atmosphere in any time for which the forecast is being made increases. point could be specified by seven features namely pressure, temperature, density, water content and velocity Weather is a continuous, data-intensive, component eastward, northward and upward. multidimensional, dynamic and chaotic process and these properties make weather prediction a big challenge (Lai, 2004). From ancient times, weather prediction has been one of the most interesting and fascinating domain of study, Scientists have been trying to forecast the meteorological characteristics using a number of methods, some of them more accurate than others. Lately, it has been discovered that data mining, can be successfully applied in this domain.

147

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2. RELATED ISSUES similar to those performed by the human brain. Artificial Weather prediction gives us the early warning about the Neural networks resemble the human brain in the possibility of whether there will be adverse weather following two ways: phenomena, the wind speed over land and also about the  A neural network acquires knowledge through sea surface situation. In many part of the world like learning. China and Japan, bad weather has caused a lot of havoc.  A neural network's knowledge is stored within For example in China, Mudslides and rock falls following interneuron connection strengths known as torrential rain in China have cut off key transportation synaptic weights. link in the southwest of the country. Bad weather plays havoc in city budget, causing disruption in decision It is an information processing paradigm that is inspired making by the government. In this part of the world, the by the way biological nervous systems, such as the brain, havoc caused by bad weather cannot be overemphasized. process information. It is composed of a huge number of In 2011 Lagos and Ibadan experienced heavy down pour highly interconnected processing elements (neurons) of rain which caused the destruction of lives and working in unison to solve specific problems. ANNs, like properties. This shows the importance of weather people, learn by example. An ANN is configured for a forecasting and the need to use appropriate techniques. particular application, such as pattern recognition or data classification, through a learning process. The artificial Hybrid techniques that use Artificial Neural Networks neuron is an information processing unit that is (ANN) and Fuzzy Logic (FL) called Neuro-Fuzzy fundamental to the operation of a neural network. There systems are a potentially powerful tool that can be used are three basic elements of a neuron model. Figure 1 to solve problems that are just too complex for shows the basic elements of an ANN (the perceptron conventional approaches. A Neurofuzzy system is a model): combination of artificial neural network and Fuzzy i. A set of synapses connecting links, each of Inference System (FIS) in such a way that neural network which is characterized by a weight or strength learning algorithms are used to determine the parameters of its own. of FIS. An even more important aspect is that the system ii. An adder for summing the input signals should always be interpretable in terms of fuzzy if-then weighted by the respective synapses of the rules, because it is based on the fuzzy system reflecting neuron vague knowledge (Abraham, 2002). In this study both iii. An activation function for limiting the Artificial Neural Network Multi-layer perceptron (MLP) amplitude of the output of a neuron. A typical and ANFIS Neuro-fuzzy models were used to predict the input-output relation can be expressed as future occurrence of rainfall in Ibadan, Oyo-State, shown Nigeria, and the performance of both methods compared.. The true power and advantage of neural networks lies in An artificial neural network is a powerful data modelling their ability to represent both linear and non-linear tool that is able to capture and represent complex relationships and in their ability to learn these input/output relationships (Hayati and Mohebi, 2007). relationships directly from the data being modeled. The motivation for the development of neural network technology stemmed from the desire to develop an artificial system that could perform "intelligent" tasks

Figure 1: Model of a perceptron

148

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

In Figure 1: For example a temperature value can be considered as “cold" and “too-cold" at the same time, with different degree of memberships. Fuzzy logic uses 3 simple steps defined as: …………… 1 i. Fuzzification: to convert numeric data in real- world domain to fuzzy numbers in fuzzy Where: domain.

Xi represents inputs to ith node in input ii. Aggregation (rule firing): computation of fuzzy numbers (all between 0.0 and 1.0) in fuzzy Wij represents weight between ith input node and jth domain. hidden node, iii. Defuzzification: convert the obtained fuzzy b represents bias at jth node, number back to the numeric data in the real world domain net represents an adder, The advantages of fuzzy logic include: f represents the activation function. i. Mimic human decision making to handle vague concepts The type of transfer or activation function affects size of steps taken in weight space. ANN’s architecture requires ii. Rapid computation due to intrinsic parallel determination of the number of connection weights and processing nature the way information flows through the network. This is carried out by choosing the number of layers, number of iii. Ability to deal with imprecise or imperfect information nodes in each layer and their connectivity. The numbers of output nodes are fixed by the quantities to be iv. Improved knowledge representation and estimated. The number of input nodes is dependent on the uncertainty reasoning problem under consideration and the modeler’s discretion to utilize domain knowledge. The number of neurons in v. Modeling of complex non-linear problems the hidden layer is increased gradually and the natural language performance of the network in the form of an error is processing/programming capability monitored (Hayati and Mohebi, 2007). Fuzzy Logic is a powerful problem-solving methodology with wide Neuro-fuzzy refers to combinations of artificial neural applications in industrial control and information networks and fuzzy logic. It was proposed by Jang processing. (1993). It means inexact reasoning, data granulation, and

computing with words. It grants an inference structure It provides a simple way to draw definite conclusions that allows the human reasoning capacities to be applied from vague, ambiguous or imprecise information. It to artificial knowledge-based structures giving meaning resembles human decision making with its ability to work for adapting linguistic strategy into control actions and from approximate data and find precise solutions. Unlike thus offers a high-level computation. Our impression of classical logic which requires a deep understanding of a the real world is pervaded by concepts, which do not system, exact equations and precise numeric values, have severely defined boundaries for example, many, tall, fuzzy logic incorporates an alternative way of thinking, much larger than, young, etc. are true only to some which allows modeling complex systems using a higher degree, and they are false to some extent as well. These level of abstraction originating from our knowledge and concepts can be called fuzzy or vague concepts. A human experience. Fuzzy Logic allows expressing this brain works with them, while computers may not do it knowledge with subjective concepts such as "very good" (they reason with strings of 0s and 1s). and "a little bit satisfied" which are mapped into exact numeric ranges An important characteristic of fuzzy logic is that a numerical value does not have to be fuzzified using only one membership function. In other words, a value can belong to multiple sets at the same time.

149

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Fuzzy sets grants means to model the ambiguity the literature. The main strength of Neuro-fuzzy systems associated with vagueness, imprecision, and lack of is that they are universal approximators with the ability to information concerning a dilemma. Consider the solicit interpretable IF-THEN rules (Wikipedia, 2008). significance of a “short person.” For an entity X, the Neuro-fuzzy modeling can be classified as linguistic short person may be one whose height is below 4.20. For fuzzy modeling that is focused on interpretability. This is another individual Y, the short person may be one whose mainly the Mamdani model, for example, Neurofuzzy height is beneath or equal to 3.9. This “short” is called as classification (NEFCLASS) and precise fuzzy modeling a linguistic descriptor. The term “short” informs the same that is focused on accuracy. This is mainly the Takagi- meaning to the individuals X and Y, but it is establish Sugeno-Kang (TSK) model, for example, Adaptive that they both do not supply a unique definition. The term Neurofuzzy inference system (ANFIS). ANFIS is an “short” would be conveyed successfully, only when a adaptive network. An adaptive network is a network of computer compares the given height value with the pre- nodes and directional links. Associated with the network assigned value of “short.” This variable “short” is called is a learning rule: for example back propagation. It’s linguistic variable, which represents the elusiveness called adaptive because some, or all, of the nodes have existing in the system. parameters which affect the output of the node. The networks learn the relationship between the inputs and Neuro-fuzzy results in a hybrid intelligent system that outputs. The ANFIS approach learns the rules and synergizes these two techniques by combining the membership functions from data. The ANFIS architecture human-like reasoning style of fuzzy systems with the is presented in figure 2. The circular nodes represent learning and connectionist structure of neural networks. nodes that are fixed whereas the square nodes are nodes Neuro-fuzzy hybridization is widely termed as Fuzzy that have parameters to be learnt. Neural Network (FNN) or Neuro-Fuzzy System (NFS) in

Layer 1 Layer 2 Layer 3 Layer 4 Layer 5

A1 w1 w1 w 1f1 X

A2  F

B1 w w w f Y 2 2 2 2

Figure 2: An ANFIS architecture for a two rule Sugeno system

A Two Rule Sugeno ANFIS has rules of the form:

If x is A1 and y is B1 THEN f1  p1x  q1 y  r1...... 2

If x is A and y is B THEN f  p x  q y  r 2 2 2 2 2 2 …………………. 3

150

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

When training the network there is a forward pass and a w f backward pass. The forward pass propagates the input O  w f  i i i vector through the network layer by layer. In the 5,i  i i i  wi backward pass, the error is sent back through the network i ………………….… in a way similar to back propagation. …………………………… 9 In layer 1, the output of each node is: ANFIS Creates a fuzzy decision tree to classify the data it O   (y) for i  3,4 is working on into one of linear regression models to 1,i Bi2 ………………... 4 minimize the sum of squared errors (SSE):

And O1,i (x) is the membership grade for x and y .

………………………………………… The membership functions could be any shape. Using the …………… 10 bell shaped function given by: 1 where:  A (x)  2b x  c i i. ej represents the error between the desired and 1 i the actual output a ii. p represents the number of fuzzy partitions of i ……………………. 5 each variable iii. n represents the number of input variables Where a ,b ,c are parameters to be learnt. These are i i i the premise parameters. 3. MATERIALS AND METHODS 3.1 Data Preprocessing In layer 2, every node in this layer is fixed. This is where The case data for the study was obtained from the the t-norm is used to ‘AND’ the membership grades, for metrological station in Ibadan and it covers a period of example the product: 125 months from January 2002 to May 2012. The preprocessing of the data was first carried. Missing O  w   (x) (y), i 1,2 values were replaced with zeros. The meteorological 2,i i Ai Bi …….. 6 dataset had seven (7) attributes; their type and description

are presented in Table 1. layer 3 contains the fixed nodes which calculates the ratio of the firing strengths of the rules: Table 1: Attributes of Meteorological Dataset Atrribute Type Description Year Numerical Year considered wi O  wi  3,i w  w Month Numerical Month considered 1 2 ………..……… 7 Min temperature Numerical Monthly Maximum temperature The nodes in layer 4 are adaptive and perform the Max temperature Numerical Monthly minimum consequent of the rules: temperature Rainfall Numerical Total monthly rainfall (mm) O  w f  w ( p x  q y  r ) 4,i i i i i i i …………. 8 Wind run Numerical Wind Run in km Relative humidity Numerical Monthly Relative Humidity in % The parameters in this layer ( pi ,qi ,ri ) are to be determined and are referred to as the consequent A graph of all the attributes was plotted to see the trend parameters. of the data using Matlab software before it was

normalized to an appropriate scale by using the Standard In layer 5 there is a single node that computes the overall Deviation and Variance using equation 11.The scaled output: data was used to train the networks.

151

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

()s U()() k U k  Uk()  NN 2  p1 p p Uk()  ……………. (11) E()()()() k E k  Y k  Yˆ k …… (13)   Y()() k Y k pp11N Yk()S ()    Yk()  Where the index p ranges over the set of input patterns

N and Ekp () represents the error on pattern . Both the where Uk( ), Yk() and Uk(),  Yk() are the mean and Back propagation of errors which is the most commonly standard deviation of the input and output training data pair; used supervised training algorithm in the multilayer feed ()S ()S Uk() and Yk() are the scaled inputs and outputs forward networks and the Levenberg-Marquardt algorithm respectively. After training the network, the joint weights were (LMA) were used and their performance compared. rescaled using equation 12. The ANFIS Model (Adaptive Neuro-fuzzy Inference System) In the ANFIS model, crisp input series are converted to ˆˆˆˆ fuzzy inputs by developing membership functions for each Y(,()) k k Y (,()) k  k Yk() Y () k … (12) input series. The Neuro-fuzzy model builds intelligence and reasoning into the system by performing Subtractive Clustering on the fuzzy sets to determine the number and With the rescaling the trained network can work with other type of fuzzy membership function. The membership unscaled validation and test data not used for training the function pattern used for the input series is of the Bell network. However, for notational convenience, shape. The goal of ANFIS is to find a model, which will ()S simulate correctly the inputs with the outputs data. The and Y()() k Y()S k were used. The 125 U()() k U k ANFIS Matlab Fuzzy Logic Tool Box was used for the data records used were divided into 100 training and 25 testing study. The ANFIS network was trained using a hybrid (validation) data sets. learning algorithm that uses least squares method in the forward pass to identify the consequent parameters of the The Neural Network Model layer 4, while in the backward pass the errors are The objective of the training of the neural network model propagated backwards and the premise parameters are is to minimize a global error that measures the difference updated by gradient descent method. Figure 4 shows the between the model output and the actual values that are screen shot of the ANFIS GUI editor. defined in equation 13.

152

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 4: ANFIS GUI Editor showing test data with the FIS Output The ANN and ANFIS models used for the analysis of predicted values. It is given by equation the weather attributes were trained for  , =100 epoch 16: each. The performance metrics used in evaluating the models were: i. Computation Time: the time taken to build the model (in seconds). …… ii. K-step ahead prediction: this is given …………………….(16) by equation 14:

4. RESULTS AND DISCUSSION ….(14) The inspection of one-step ahead output Training and validation result for the neural network predictions might not reveal the model models inaccuracy for multi-step output or The training and validation results of the neural network distant predictions. models using the Back propagation of errors and iii. Mean Squared Error: This is one of Modified Levenberg Marquardt Algorithm (MLMA) are the most commonly used performance presented. Figure 6 shows the training result for the metrics. It is computed by taking the minimum temperature; figure 7 shows the validation average of the squared differences result for the minimum temperature, figure 8 shows the between each computed value and its training result for the maximum temperature, figure 9 corresponding correct value given by shows the validation result for the maximum temperature, equation 15: figure 10 shows the training result for the Rainfall data, figure 11 shows the Validation result for the Rainfall data, figure 12 shows the training result for the Wind data, ……(15) figure 13 shows the validation result for the Wind data, iv. Root Mean-Squared Error: This is figure 14 shows the training result for the Relative simply the square root of the mean- Humidity data and figure 15 shows the Validation result squared-error. The mean-squared error for the Relative Humidity Data. gives the error value the same dimensionality as the actual and

Figure 6: Training result for the minimum temperature data

153

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 7: Validation result for the minimum temperature

Figure 8: Training result for the maximum temperature data

Figure 9: Validation result for the maximum temperature

154

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 10: Rainfall training result

Figure 11: Validation result for the rainfall data

Figure 12: Wind training Result

155

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 13: Validation result for the Wind data

Figure 14: Relative Humidity Training Result

Figure 15: Validation result for the Relative Humidity data

156

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

It can be seen that the network prediction of the training data and validation data of the MLMA generally match closely the original training data. This shows that the trained network and the validation network generally captures and approximate the system accurately as indicated by the value of the mean square error presented in table 2.

Table 2: Training and Validation Result For the Neural Network Algorithm Weather Computation Minimum Mean Squared Mean Squared KSTEP Ahead attribute time performance Error Error (Prediction Error

index (Validation MSE) MSE)

BP MLMA BP MLMA BP MLMA BP BP MLMA

MLMA

Min Temp 1.966 1.591 0.129 0.000 -0.0902 -0.000 0.0284 0.000 -6.857 -8.73

Max Temp 2.090 1.872 0.703 0.000 4.164 0.000 0.467 0.000 7.172 -7.24

Rainfall 2.168 1.685 0.571 0.006 283.053 3.976 0.069 -0.000 -572.96 -395.35

Wind 2.122 1.997 0.697 0.000 -192.59 0.3611s -0.017 0.000 -59.86 -34.96

Relative 2.574 1.700 0.516 0.003 -2.736 0.663 -0.035 0.000 -12.76 -9.43

Humidity

4.1 The ANFIS Model Result Using the Matlab GUI Text viewer the following results were obtained for the ANFIS network: 1. Number of nodes: 204 2. Number of linear parameters: 100 3. Number of nonlinear parameters: 150 4. Total number of parameters: 250 5. Number of training data pairs: 100 6. Number of checking data pairs: 25 7. Number of fuzzy rules: 50

Figure 16 shows the ANFIS training result for the minimum temperature data, figure 17 shows the ANFIS validation result for minimum temperature data, figure 18 shows the ANFIS training result for maximum temperature data, figure 19 shows the ANFIS validation result for maximum temperature data, figure 20 shows the ANFIS training result for rainfall data, figure 21 shows the ANFIS training result for rainfall data, figure 22 shows the ANFIS training result for wind data, figure 23 shows the ANFIS validation result for wind data, figure 24 shows the ANFIS training result for relative humidity data and figure 25 shows the ANFIS validation result for relative humidity data.

157

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 16: ANFIS Minimum Temperature Training Result

Figure 17: ANFIS Minimum Temperature Validation Result

158

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 18: ANFIS Maximum Temperature Result

Figure 19: ANFIS Validation result for the Maximum Temperature data

159

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 20: ANFIS Rainfall Training result data.

Figure 21: ANFIS Rainfall Validation result data.

160

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 22: ANFIS Training result for Wind Data

Figure 23: ANFIS Validation result for the Wind data

Figure 24: ANFIS Training Result for the Relative Humidity data

161

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Figure 25: ANFIS Validation Result for the Relative Humidity data

It can be seen that the network prediction of the training data of the ANFIS and validation data also closely match the original training data. This also shows that the trained network and the validation network generally captures and approximate the system accurately. The Mean square error can be seen in table 3. Table 4 presents a comparison of the best neural network and ANFIS models.

Table 3: Result of the ANFIS Model Weather Minimum MSE RMSE KStep Ahead attribute performance index Prediction Error Min Temp 0.0085 0.0133 0.1153 0.1059 Max Temp 0.00069 0.0118 0.1086 0.0989 Rainfall 0.4566 1.0185 1.0092 6.9522 Wind 0.0017 0.002 0.0447 3.4124 Relative Humidity 0.0150 0.0271 0.1645 0.3067

Table 4: Comparison of the ANN and Neuro-fuzzy models performance ANN (MLMA) ANFIS (MLMA) Weather attribute MSE RMSE MSE RMSE Minimum Temperature -0.000 0.000 0.0133 0.1153 Maximum Temperature 0.000 0.000 0.0118 0.1086 Rainfall 3.976 1.993 1.0185 1.0092 Wind 0.3611 0.6009 0.002 0.0447 Relative Humidity 0.663 0.8142 0.0271 0.1645

4. DISCUSSION OF THE RESULT

From the comparison of the two models used in this Both the neural network and ANFIS MLMA trained study both the neural network model and the ANFIS networks performed better than their corresponding neurofuzzy model were able to capture the dynamic Back propagation trained networks. Of the two models behavior of the weather data, resulting in a more the ANFIS model gave a better performance. compact and natural internal representation of the Temperature, Rainfall, Wind and Relative humidity information contained in the weather profile.

162

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFRENCES Bharati, M. Ramageri (2010) Data Mining Techniques And Applications Indian Journal of Abraham A., (2002). Evo NF A frame work for Computer Science and Engineering Vol. 1 optimization of fuzzy inference systems No. 4 301-305 using Neural Network Learning and th Evolutionary computation. The 17 IEEE Casas, D. M, gonzalez a.t, rodrígue j. E. A., pet j. V., International Symposium on Intelligent (2009) Using data-mining for short-term control ISIC 02,IEEE Press pp.327-332,2002 rainfall forecasting, notes in computer science, volume 5518, 487-490 Jang, (1993), ANFIS: adaptive network-based fuzzy inference systems, IEEE Transactions on Ddamousis, I. G., and Dokopoulos, P. (2001), A fuzzy Systems, Man and Cybernetics, vol. 23, no. expert system for the forecasting of wind 3, pp. 665 – 685. speed and power generation in wind farms, proceedings of the second ieee international Lai, (2004). Intelligent weather forecast. Third conference on power industry computer international conference on machine learning applications, sydney australia. 63-69. and cybernetics, shanghai. Dennis, J. E. , Schnabel, R.B. (1996). “Numerical Hayati M. and Mohebi Z., (2007). Application of Methods for Unconstrained Optimization and Artificial Neural Networks for Temperature Nonlinear Equations”.. Englewood Cliffs, Forecasting. World Academy of Science, NJ: SIAM, Prentice-Hall. Engineering and Technology Domańska, D., (2002). Fuzzy weather forecast in Wikipedia, (2008), "climate change" from wikipedia - forecasting pollution concentrations. the free encyclopedia, retrieved from Department of modelling and computer http://en.wikipedia.org/wiki/climate_change graphics, institute of informatics, university in august 2008 of silesia, b¸edzińska 39 sosnowiec, 41-200, poland [email protected] m. Wojtylak institute of meteorology and water BIBLIOGRAPHY management (imgw), bratków 10 katowice, 40-045, Poland Abraham A & Nath b. (2000), Designing optimal [email protected] neuro-fuzzy systems for intelligent Control. the sixth international conference on control, Elia G. P. (2009), A decision tree for weather automation, robotics and vision, Icarcv Wang prediction , universitatea petrol-gaze din J L (editor), (cd rom proceeding, paper ploiesti,bd. Bucuresti 39, ploiesti, catedra de reference 429 - fp7.3 (i) informatică, vol. Lxi, no. 1

Abrahan A., (2004).Intelligent Systems Architectures Hagan, M. T. and Menhaj, M. B. (1994). Training and Perspectives. Faculty of Information feedforward network with the Marquardt Technology, School of Business Systems algorithm . IEEE Trans. Neural Netw., vol. 5, Monash University (Clayton Campus), no. 6, pp. 989 – 993. Victoria 3168, Australia Email: [email protected], URL: Hansen, B.K., D. Riordan, (2001) .weather prediction http://ajith.softcomputing.net using case-based reasoning and fuzzy set theory, in: proc. Workshop on soft computing Adeli, H. & Jiang, X. (2002). Neuro-fuzzy logic model in case-based reasoning, international conf. for free-way work zone capacity estimation. On case-based reasoning, canada, , pp.175- J. Transportation Engineering, Vol. 129, pp. 178. 484-493. Hung, N.Q, Babel, M.S, Weesakul, S. Andtripathi, Alex G. BÜCHNER, Maurice D. MULVENNA, Sarab N.K.(2008).An artificial neural network for S. ANAND, John G. HUGHES An Internet- rainfall Forecasting in Bangkok ,Thailand. enabled Knowledge Discovery Process. Northern Ireland Knowledge Engineering James W. Taylor, Roberto Buizza(2002).Neural Laboratory χ School of Information and Network Load Forecasting with Weather Software Engineering Faculty of Informatics Ensemble Predictions . IEEE Trans. on University of Ulster. Power Systems, 2002, Vol. 17, pp. 626-632.

163

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Jang J. S. R., (1993), ANFIS: Adaptive-network-based engineering and automation Instituto superior fuzzy inference systems. IEEE Transaction de engenharia de lisboa R. Conselheiro Systems, Man and Cybernetics, Vol. 23, pp. emídio navarro, 1950-062 lisbon (portugal) 665-685. [email protected]

Jyosthna Devi , B.Syam , Prasad Reddy, K.Vagdhan Smail Khalifa, Mohammad Hassan Komarizadeh Kumar ,B.Musala Reddy,N.Raja , Nayak (2010). An intelligent approach based on (2004). ANN Approach for Weather adaptive neuro-fuzzy inference systems Prediction using Back Propagation, (ANFIS) for walnut sorting, Department of Department Of Computer Science and Mechanical Engineering of Agricultural Engineering, KLCE, Vaddeswaram, Guntur Machinery, Urmia University, Urmia, Iran, Dt.-522502, Andhra Pradesh, India. Corresponding author: [email protected] Mathworks, Fuzzy Logic Toolbox – ANFIS and the ANFIS Editor GUI, MATLAB 7.0.1 Tabesh, M. (2006). Short-term water demand prediction using neural networks and neuro- Mirmomeni, Caro Lucas, Behzad Moshiri , Babak fuzzy systems. Final Report, University of Nadjar Araabi” (2010) Adaptive Neurofuzzy Tehran, Water Institute, Ministry of Energy, modelling with online learning method for Iran. predicting time varying solar and geomagnetic time activity indices. Masoud Wang, X. G., Tang, Z., Tamura, H., Ishii, M. and Sun, Journal expert system with application :An W. D. (2004). “An improved international journal archive volume 37/ backpropagation algorithm to avoid the local minima problem” Neurocomputing, vol. 56, Negnevitsky, M. and C. W. Potter, (2006) Innovative pp. 455 – 460. Short-Term Wind Generation Prediction Techniques, in Power Systems Conference Wansalfarin Abintiwan Husain (2008) Fuzzy case- and Exposition. 2006, pp. 60-65. based reasoning for weather prediction. University Teknologi, Malaysia, October Nelles, O. (2001). Nonlinear system identification from 2008. classical approaches to neural networks and fuzzy models.. Springer Publisher. New Yu, D. L., Yu, D. W. and Gomm, J. B. (2006). “Neural Jersey: Prentice Hall, 1997, pp. 73,74, 86, model adaptation and predictive control of a 95-97,86-87, 26-28, 74-85.. chemical process rig”. IEEE Trans. Cont. Sys. Tech., vol. 14, no. 5, pp. 828 – 840. Pousinhol, H.M.I., V.M.F. Mendes J.P.S. Catalão (2001) Neuro-fuzzy approach to forecast wind power in Portugal. Department of electromechanical engineering University of beira interior R. Fonte do lameiro, 6200-001 covilhã (portugal) [email protected], [email protected] department of electrical

164

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Web Designers Guide on Development Technologies: An Evaluation Approach

1F.E. Ekpenyonga & D.T. Chinyio Department of Mathematics and Computer Science, Nigerian Defence Academy Kaduna. Email: [email protected], Tel N0 08030802228

1Correspondence Author

ABSTRACT The advent of network computing has brought about a lot of technological innovations among which are web development tools. Selecting a suitable one has proven to be difficult. The purpose of this study is therefore concerned with a survey of some popular web development technologies by embarking on an intensive evaluation exercise. Some evaluation techniques were adopted and implemented on each of the selected technologies by using both primary and secondary data. A wide range of interview was conducted, an empirical approach also used to evaluate the Web development tools using some decision variables such as Platform, Speed of execution, Database supports, Program length etc. A usability study was embarked through the use of the distribution of structured questionnaires to target respondents which include experienced programmers, Web designers, Computer IT training Institutions with carefully selected experimental. The data were statistically analysed with the use of Mean Comparison of Web Technologies against the variable matrices and useful results were derived. It was apparent from the outcome of the result that the productivity of software is partially determined by such factor as web development tools, machine speed, and other factors.

Keywords: Web Technology, Browsers, Internet, Primary and Secondary data. . African Journal of Computing & ICT Reference Format F.E. Ekpenyonga & D.T. Chinyio (2013 Web Designers Guide on Development Technologies: An Evaluation Approach. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 165-172

1. INTRODUCTION 1.1 Background To Research Study The Internet is no longer merely an e-mail and file- The internet and the World Wide Web as it stands today sharing system as it was originally intended, but rather is a magnificent success, as an informational has emerged as a prevailing interactive standard. It has superhighway aimed at stifling the limitations of grown to be an integral part of the ever-escalating large- conventional communication means, and of allowing the scale media system, moving into center stage of media communication between humans and machines. It has politics alongside traditional broadcast media television fast become a fundamental source of information to and radio. All the credits to the technology of the World almost every facet of life, and as a cradle of learning, Wide Web which has now become a major delivery research, recreation [5]. The main goals of evaluation are platform for web development a variety of complex and to assess the application functionality, to verify the effect sophisticated enterprise applications in several domains. of its interface on the user, and to identify any specific These web applications exhibit complex behavior and problem with the application, such as aspects which show place some unique demands on their usability, unexpected effects when used in their intended context performance, security and ability to grow and evolve. [17]. However, a vast majority of these applications continue to be developed in an ad-hoc way, contributing to Evaluating Web applications in particular entails problems of usability, maintainability, quality and verifying if the application design allows users to easily reliability[18]. retrieve and browse contents, and invoke available services and operations. This therefore implies not only 1.2 Statement Of The Problem having appropriate contents and services available into As the technology of the web expands, its social the application, but also making them easily reachable by recognition and audience grows. The challenges of the users through appropriate hypertexts. The interest of this web developers remains as it has been for years, to value research is to explore some of the popular Web and significance to the online medium. Professionally development tools that are mostly used by web developed websites involve major investment of talents developers in Nigeria. and resources.

165

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The people funding this works naturally begin to enquire The main deficiencies in web development tools are that what kind of returns they are getting as very few of them they cannot support teams of developers working have much of a feel for their reimbursement on this together and are incapable of building flexible venture. Much of that has been due to the incredible hype applications that can be adapted to meet the changing and fast growth surrounding this technology, combined business and technology requirements of different with the low cost of experimentation with the latest and organisations [8],[19]. According to [9],[18],Web emerging sophisticated Web development tools or development tools evolve extremely fast, enabling technologies available in the software market that suit sophisticated tools to be deployed and complex their need. interactions to take place. It is a well known fact that the average quality of Web sites is poor due to “lack of Since software engineering techniques have not sophisticated Web development technologies that progressed at a sufficiently rapid pace to cover the enhances the dynamic and interactivity of Web pages specifics of web application and to cope with upcoming being the main cause of user dissatisfaction”. technologies, any new ideas and discoveries are never wholly accepted without subjecting them to some criteria Other citations on web developments could be seen in the through research work. Web design and maintenance are works of [10],[11],[21],[22],[23]. Thus, Web likely to absorb more and more resources as web development technologies quality can be defined technologies and users keep evolving at an alarming rate. (ISO9126) as “the totality of features and characteristics of a software product that bear on its ability to satisfy 1.3 Study Objectives stated or implied needs” such as robustness, reliability, This research work evaluated some web development maintainability, and usability as the main significant technologies. Therefore the investigation carried out on factors. Also, usability can be defined (ISO9241) as “the the following specific areas: effectiveness, efficiency and satisfaction with which i. Identify personal characteristics of the specified users achieve specified goals in particular respondents and the extent of knowledge about environments”[16]. web development tools ii. Access the perceived usability of the given 2.1 Web technologies tools over some period of time Web technologies are the programming languages, iii. Identify the most preferred tools in the study document display principles and services existing when areas using the Web to relate with the Internet.

2. RELATED LITERATURE 2.1.1 The Types of Web Technology Web development tools are failing to address users’ Several scores of web technologies are handy today. needs despite the promises made by vendors. Users are Ranging from communication services to analysis tools presented with the opportunity of using the Web to for business marketing. Some of the key technologies that decrease overheads associated with client/server allow us to explore the Internet are as follows: applications, to create new sales channels and facilitate  Markup languages the information flow between customers and  Web services suppliers[15]. The Web development tools market is  Web analytics immature, with many different methods and approaches on offer, and several conflicting standards. However, the 2.1.2 Markup Languages market has no clear leader, and there are inadequate Markup languages are a standard coding or language products for medium or large-scale development projects interpreted by Internet browsers. They are designed to [19]. allow you create electronic documents that are most commonly formatted for display on the computer screen Also evaluated a wide cross-section of Web Development in a Web browser. One of the most popular markup Tools based on Vendors’ approach by taking a critical languages is Hyper Text Markup Language, or HTML. scrutiny of eight most popular vendors’ such as: This language was developed in 1990 by Tim Berners- Netscape, Oracle, IBM, Vision Software, Passport Corp, Lee. Sybase, Intelligent Environments, Microsoft using some identified issues such as scalability, application Other examples of markup languages include the adaptability, deployment flexibility, Change-cycle following: support, platform compatibility, and database Extensible Markup Language (XML) compatibility. He explains the strengths and weaknesses Extensible HTML (XHLML) of each. He concluded that, it saves you time and money selecting the right strategy and tool.

166

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

XHTML is HTML as defined through XML (Extensible available in other web authoring tools. However, it is Markup Language). XML allows progammers who important to remember that some of these functions will understand it to write their own markup commands, or only work when your website has been uploaded to a modify current ones, increasing the flexibility of a server that supports Microsoft Front Page server language such as XHTML[8]. extensions.

2.1.3 Web Services Dreamweaver Adobe Dreamweaver is a web Web services make the Internet a vibrant place. They development tool originally owned by Macromedia. It services enables devices, such as computers and mobile supports CSS, JavaScript, and a number of other web- equipments, to communicate over the Web. Many related technologies. It combines HTML and CSS plain services are provided in an obscure way to the user. For text editing with a WYSIWYG tool which lets you check example, online storage and other services are offered as progress as you go or edit WYSIWYG on the fly. It's the a background work by cloud computing. industry standard tool for this sort of thing, although there are many, many programs which do more-or-less 2.1.4 Web Analytics the same thing.CGI. Dreamweaver is not designed to Web analytics involves the collection of complex data completely remove the agency of HTML and CSS: it is from an Internet domain name and twisted into reports by meant to assist you with your HTML and CSS and to the domain owner. The reports provide detailed make some of the more mundane aspects of Web information about a number of events. For example, Publishing less terrible[22]. Online auction-sites such as eBay may use web analytics to determine which part of a website results in the most CGI stands for Common Gateway Interface. Similar to sales. ASP and PHP, CGI is used for server-side processing for Web applications. Because CGI is designed to be server- 2.2 Web development agnostic, you can develop CGI applications that run on Web development is an extensive term for the process of Windows, UNIX, Macintosh, or other server operating creating a web site for the Internet (World Wide Web) or systems. You can write CGI applications in C, C++, Java, an intranet (a private network). This can include web and Perl. ASP is an alternative to traditional CGI design, web content development, client liaison, client- programming. side/server-side scripting, web server and network security configuration, and e-commerce development. ColdFusion was introduced in 1995 by Allaire Corporation as an “easy to program” technology for 2.2.1 Some selected web development tools designing and creating dynamic content. The vital feature PHP is an open-source, cross-platform, object-based of the technology is the template-based ColdFusion scripting language for vibrant Web applications that is Markup Language (CFML), originally interpreted but equivalent to ASP. Versions of PHP run on numerous now JIT-compiled into servlets. It is a scripting language operating systems and Web servers, and interfaces are that has been used by developers to create dynamic available to many database systems. PHP is available for websites. A dynamic website can alter depending on Windows operating systems, but since ASP being outside factors like data, user preferences or changes in commonly the preferred alternative on Windows systems, backend database. Websites that only use HTML are PHP is most prevalent on Linux systems running Apache. static.

HTML stands for Hyper Text Markup Language, ASP is a server-side scripting environment for Web founded in 1980 by Tim Berners-lee. It provides a means applications that provides a scripting language engine to create structured documents by denoting structural that supports several languages, an object model, and an semantics for formatting documents. It enables images interface for accessing server components. Although ASP and objects to be embedded and can be used to create is generally used with VBScript, you can also use ASP interactive forms. It is written in the form of HTML with JavaScript. When used with VBScript, you can use elements consisting of "tags" surrounded by angle ASP for database processing, form processing, and other brackets within the web page content [15]. Web applications that require server interaction, such as sending mail and reading or changing the contents of files FrontPage Microsoft Front Page is a web page that are located on the server. authoring tool designed to fully integrate page editing, site management, reporting, and publishing within the XML The Extensible Markup Language (XML) is a same application. Front Page is what is known as a widely accepted markup language that simplifies the WYSIWYG ("what you see is what you get") editor, transmission of structured data between applications. much like other editors like Dreamweaver or Composer, XML is a meta-language for creating collections of which allows you to forego actual coding and simply custom elements, in contrast to HTML, which provides a design WebPages within the document window. Front fixed set of elements. Page offers many unique functions and features not

167

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

XHTML stands for eXtensable HyperText Markup With flash the web developer is now able to create a user Language and is a hybrid of HTML and XML. XHTML experience that is rich in media and relatively quick was created for two main reasons: To create a stricter loading, especially compared to traditional methods like standard for making web pages, reducing GIF animations incompatibilities between browsers and to create a standard that can be used on a variety of different devices Web Browser without changes is a web standard which has been agreed A web browser is a software application for retrieving, by the W3C and, as it is backwards compatible presenting, and traversing information resources on the World Wide Web. An information resource is identified FLASH Macromedia Flash, otherwise called Adobe by a URL and may be a web page, image, video, or other Flash or Shockwave Flash, is a popular program used in piece of content. create animations, advertisements, web pages and Primary tasks of a web browser is to Convert web presentations. It was introduced in 1996. Macromedia addresses (URL’s) to HTTP requests, Communicate with Flash contains a scripting language called Action Script. web servers via HTTP and to finally Render The program Flash was the ingenuity of Jonathan Gay, (appropriately display) documents returned by a server. whose initiative started in college and developed the Hyperlinks present in resources enable users to easily software while working for Silicon Beach Software. navigate their browsers to related resources. Although Macromedia Flash itself can create a whole webpage and browsers are primarily intended to access the World it is livelier than plain HTML. Wide Web, they can also be used to access information provided by web servers in private networks or files in file systems. The major web browsers are Windows Internet Explorer, Mozilla Firefox, Apple Safari, Google Chrome, and Opera[6].

A Recent Browser Use Statistics as shown in the figure shows that Chrome and Firefox are the most patronized browser as reflected in the table below:

Table 1: Web Browsers’ User Statistics

2012 Internet Explorer Firefox Chrome Safari Opera October 16.1 % 31.8 % 44.9 % 4.3 % 2.0 % September 16.4 % 32.2 % 44.1 % 4.2 % 2.1 % August 16.2 % 32.8 % 43.7 % 4.0 % 2.2 % July 16.3 % 33.7 % 42.9 % 3.9 % 2.1 % June 16.7 % 34.4 % 41.7 % 4.1 % 2.2 % May 18.1 % 35.2 % 39.3 % 4.3 % 2.2 % April 18.3 % 35.8 % 38.3 % 4.5 % 2.3 % March 18.9 % 36.3 % 37.3 % 4.4 % 2.3 % February 19.5 % 36.6 % 36.3 % 4.5 % 2.3 % January 20.1 % 37.1 % 35.3 % 4.3 % 2.4 % Data Source: W3Schools.com

Client/Server side Scripting How they Communicate Client side coding such as XHTML is executed and Initial Client Request: The HTTP client sends a request stored on a local client (in a web browser) whereas message formatted according to the rules of the HTTP server side code is not available to a client and is standard. This message specifies the resource that the executed on a web server which generates the client wishes to retrieve, or includes information to be appropriate XHTML which is then sent to the client. provided to the server. Response by Server: The server reads and interprets the request. It takes action relevant to the request and creates an HTTP Response message, which it sends back to the client. The response message indicates whether the request was successful, and may also contain the content of the resource that the client requested, if necessary.

168

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Web-Server Our target respondents include experienced The term web server or webserver can mean one of two programmers, Web designers, Computer IT training things: A computer program that is responsible for Institutions etc. A statistical analysis was done on the accepting HTTP requests from clients (user agents such data using the Mean Comparison which gives a better as web browsers), and serving them HTTP responses result than a mere Percentage analysis used by some along with optional data contents, which usually are researchers. The decision variables used are as follows: web pages such as HTML documents and linked  Database supports objects (images, etc.). Or a computer that runs a  Portability computer program as described above. Its basic  Program length responsibility is to Receive HTTP request via TCP,  Speed of execution Map Host header to specific virtual host(one of many  Platform host names sharing an IP address), Map Request-URI to specific resource associated with the virtual host and The choice of the decision variables is guided by the return the HTTP request and response. a recent browser fact that the speed of a technology is highly dependent statistic is shown if fig 2 on the chosen variables.

Table 2: Demographic status of the respondents. s/n Personal Freq Perc (%) Characteristic s SEX Male 98 76.5625 Female 30 23.4375

EDU STATUS SSCE 20 15.625 OND 32 25 HND 27 21.09375 Fig 1: recent server statistics. NCE 1 0.78125 1st Degree 37 28.90625 3. METHODOLOGY Masters 11 8.59375

Depending on the phase in which evaluation is AGE 20-29 35 27.34375 performed, it is possible to distinguish between 30-39 57 44.53125 formative evaluation, which takes place during design, 40-49 32 25 and summative evaluation, which takes place after the 50 And Above 4 3.125 product has been developed, or even when any prototype version is ready. During the early design EMPL stages the goal of the formative evaluation is to check STATUS Student 51 39.84375 the design team understanding of the users’ Private Sector requirements, and to test design choices quickly and Empl 65 50.78125 informally, thus providing feedback to the design Civil Servant 12 9.375 activities. Later on, the summative evaluation can support the detection of users’ difficulties, and the The usage frequency of Web development technologies improvement and the upgrading of the product. [17]. from 128 participants was identified (see Table 3.1). There were 76.5625% male and 23.4375% female We adopted the empirical approach to evaluate the respondents. Their educational status 15.625% School Web development tools. In tying make a fair judgement certificate holders, 25% Ordinary Level Diploma in coding, compiling / interpreting and running certificate holders, 21.09375% Higher National programs in each Web tools, a sample web page was Diploma degree holders 0.78125% NCE, 28.90625% designed and tested against each variable. The Web have University Degree and 8.59375% have Masters development technologies considered include: PHP, Degree. On the age category, about 27% were in their HTML, MS-FrontPage, VBScript, Active Server Page, twenties, 45% in the grade of thirties 25% were in the JavaScript, Macromedia Dreamweaver, Macromedia 40yrs age grade while 3% were above the stated age ColdFusion, XML, Macromedia Flash, jQuery. A grades. About 40% were students, 51% in the private usability study was also deployed in the research which sector and only 9% were Civil servants. involved the design and distribution of structured questionnaires to target respondents.

169

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4. RESULTS AND FINDINGS

TABLE 3: Mean Comparison of Web Technologies against the variable matrices (speed)

VARIABLES * TECHNOLOGIES Cross tabulation

Count

TECHNOLOGIES Means

PHP HTML FP DW CF ASP JVSC VBSC XML FLASH

VARIABLES PF SE PL 16.00 32.00 30.00 28.00 27.00 32.00 16.00 26.00 23.00 23.00 25.30 DB 94.00 85.00 98.00 87.00 86.00 88.00 82.00 94.00 91.00 95.00 90.00 PORT 58.00 74.00 40.00 35.00 77.00 79.00 62.00 68.00 66.00 27.00 58.60 Means 80.00 24.00 18.00 25.00 63.00 74.00 67.00 73.00 79.00 28.00 53.10 93.00 71.00 64.00 80.00 89.00 86.00 89.00 81.00 92.00 72.00 81.70 68.20a 57.20a 50.00a 51.00a 68.40a 71.80a 63.20a 68.40a 70.20a 49.00a 61.74a

Note: Figures followed by the same letter are statistically not significantly different at 5% significant level according to Least Significant Different (LSD) Test.

The table above shows that there is no significant difference in the overall comparison of the development technologies. However, PHP, FLASH, FrontPage, XML, VBScript have the high execution speed. PHP, CF, JVSC, XML are favoured in terms of portability. PHP JVSC, XML, FLASH, PHP have the less dependent on the platforms. PHP ASP JVSC XML all have high support for database.

Figure 2: Graph of Technologies Against Mean Values.

170

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4.1 Result Discussion: Web-applications are to some extent still developed as From the chart of Mean Comparison of Web if they were common tool mainly used for publishing Technologies against the variable matrices the information. The significance of careful considerations following deductions are made; PHP, FrontPage, of design and architectural choices needs to be HTHL, have great execution speed. Front Page, understood by the involved parties, communicated by Dreamweaver, Macromedia Flash exhibit low tendency the key actors in the actual development efforts in to support vast databases. Almost all the technologies getting well acceptable design. Hence, a main show relative high level of portability. In terms of challenge for headway of web application is a complete program length, HTML, Coldfusion, ASP, VBScript transformation in the applied practices. HML have longer lines of codes as compared to the others. Acknowledgement Now, a comparison of the result from Table1 and I wish to acknowledge the contribution of Professor figure1 clearly shows that the preferred tools in terms Bamidele Oluwade for his technical support and of speed are PHP,XML,ASP. supervision of this research work.

5. SUMMARY OF FINDINGS: References: It is observed from the findings that a great deal of web [1] Active Server Pages : "An Introduction to application development is vastly motivated by Web-based Application Development" technology drive in high speed inventive scenery. available at Corresponding with the shifts in importance of the web http://www.abiglime.com/webmaster/articles/ applications developed, we anticipate a shift towards asp/122297.htm extra technology driven approaches where needs, [2] Amadin I.F.(2010) An Empirical Comparison requirements, and high quality becomes fundamental. Of: HTML, PHP,COLDFUSION, PERL, However, despite the fact from our findings that there ASP.NET, JAVASCRIPT, VBSCRIPT, is no significant difference in the overall comparison of PYTON AND JSP Global Journal of the development technologies when all the factors are Computer Science and Technology Vol. 10 taken into consideration, there are still significant Issue 12 (Ver. 1.0) differences in variables when independently considered [3] Bakken S.S., (2000), “Introduction to PHP”, which brings us to the following: available at http://www.zend.com  If you are designing generally with no /zend/hop/tas,as.php specific interest in a particular variable, [4] Bakken S.S., (2000), “A Brief History of choose any technology that you are most PHP”, available at http://www.php.net/ familiar with. manual/en/intro-history.php  If your application needs support for vast [5] Barry D, Cristina V,(2005), “Survey of database, then PHP, ASP, JVSC, Technologies for Web Application XML,VBSC would be a better choice. Development” ACM Journal Name, Vol. 2,  PHP FP, VBSC, XML, FLASH all showed No. 3, June 2005, Pages 1–43. remarkable speed in execution and as such [6] Bloor Research Group: Web Based and would be ideal for designers with heavy sites Client Side Development tools: An that desire fast execution/display of their web Evaluation & Comparison available at sites. http://dpu.se/bloweb-e.html-27k [7] Bos, B., C_ elik, T., Hickson, I., and Lie, H.  JVSC, XML, FLASH, PHP have the less W. 2004. Cascading style sheets, level 2 dependent on the platforms. So it would be a revision 1, CSS 2.1 specification: W3C good choice if the designer needs the candidate recommendation 25 February application to run on different platforms. 2004. W3C Recommendation. http://www.w3c.org/TR/CSS21/. [8] Deitel, H.M., Deitel, P.J., and Goldberg, 6. CONCLUSION A.B., (2004), “INTERNET & WORLD WIDE WEB How to program”, 3rd Edition, From our research, we observed that web-applications Published by Pearson Education Ltd., are becoming vital to field of endeavors and the Singapore. demand for a higher quality of web-applications is [9] Dix, A., Finlay, J., Abowd, G., Beale, R.: nd increasing significantly. Invariably, this is not reflected Human-Computer Interaction, 2 edn in the approaches taken, the methodologies applied, (Prentice Hall, London 1998) and the organization of the work used in much [10] Felming, J., (1998), “Web navigation: development of web applications nowadays. designing the user experience”, O’ Reilly Publishers.

171

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

[11] Getting started with Perl and CGI available at [21] Ovum Evaluates: Web Development Tools http://www.perl.com/cgi available at http://www.dpu.se/o vuweb- [12] Getting started with Microsoft FrontPage e.html-32k available at http://www. [22] Macromedia Dreamweaver MX repository microsoft.com/library/en-us/frontpage.html available at http://www.macrom [13] Getting started with JavaScript available at edia.com/software/dreamwever. http://www.javaScript.com [23] Macromedia Flash MX repository available [14] Gilson, S., (2002), “Developing ColdFusion at http;//www.macromeida.com/ MX Aplications with CFML”, available at software/flash http://www.macromedia.com/php/cioldfiusio n/documentation/fmx-dev-cf-apps.pdf. [15] HTML and XHTML Tutorial available at http://www.webr eference.com/xm/ reference/xhtml.html [16] ISO (International Standard Organization). ISO 9241: Ergonomics Requirements for Office Work with Visual Display Terminal (VDT)-Parts 1-17 (1997) [17] Jeske, D., (2005), “Clearsilve compared: VS PHP, ASP, JSP” available at http://www.clearsilver.net

[18] Kantner, L., Rosenbaum, S.: Usability Studies of WWW Sites: Heuristic Evaluation vs. Laboratory Tes ting. In: ACM SIGDOC’97, International Conference on Computer Documentation, Snowbird, USA (ACM, New York 1997) pp 153-160 [19] Nielsen, J., (1999), “Designing Web usability: the practice of simplicity”, New Riders Publishing. [20] Nielsen, J., and Mack, R., (1994), 3rd Edition, “Usability Inspective Methods”, Wiley Publishers.

172

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Adaptive Estimation of Low Frequency Noise in Impedance Cardiography

Ms. Madhavi Mallam Department of ECE, DVR Kanchikacharla, India [email protected]

A. GuruvaReddy Dr.HS MIC College of Technology Kanchikacharla, India [email protected]

ABSTRACT Adaptive systems are employed in the cancelation of noises and estimation of periodic and quasi periodic signals. Amongst these signals are the electrocardiogram (ECG),impedance cardiography, brain evoked potentials and modulated signals in telecommunication applications. In this paper we study the behaviour of the weights of the LMS algorithm when used to estimate the coefficients of the discrete Fourier transform (DFT) of a signal under influence of low frequencies. It uses a reference signal, closely related to the respiratory artifact, obtained by a least-squares approximation based B-spline fitting on the contaminated imedence cardiogram in synchronism with the respiratory phases. We show theoretically that low frequency noise affects the estimation of the weights at higher frequencies. The simulation results obtained are in agreement with theoetical results.

Keywords- Adaptive algorithms, LMS algorithm, Low frequency noise. . African Journal of Computing & ICT Reference Format F.E. Ekpenyonga & D.T. Chinyio (2013 Web Designers Guide on Development Technologies: An Evaluation Approach. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 173-178

1. INTRODUCTION

Adaptive systems and estimation of periodic and quasi This method was also applied by [7] for estimating the periodic signals. Due to the noise presence in the quasi discrete Fourier coefficients of sinusoidal signals with periodic signals, the usage of adaptive filters remained known arbitrary frequencies in additive noise. Besides challenging. For example, respiration and motion adaptive spectral analysis this technique can be used to artifacts cause baseline drift in the signal waveform, filter several types of noises [1,8]. For example, it was mainly during exercise, and this drift results in suggested the scaled FLC (SFLC)in [8] to eliminate not estimation errors. Ensemble averaging of the signal only non-correlated noises but also body movements suppresses motion artifacts is used but it introduces [9]. distortion in the estimation of the signal [1]. Those interference signals have low frequency (0.04–0.15Hz) and very low frequency (0.0033–0.04Hz) components, with consequences in the analysis of the signals.

To illustrate we show in Fig. 1, the impedance cardiography(ZCG) signal. The discrete Fourier transform(DFT) and its variants are widely used in spectral analysis of biomedical signals. An adaptive method to estimate the DFT coefficients was propose by [2]. The method used the LMS algorithm [3,4] in the adaptation process. This technique, known as the

Fourier linear combiner(FLC), was later employed by [5,6]. In this application, the reference inputs are Fig. 1. FFT of an ICG. In (a) we have the ICG sinusoidal and co-sinusoidal functions and the LMS signal and in (b) the ECG signal. algorithm is used to update the weights, the Fourier coefficients.

173

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

In this study, we used the FLC to estimate the spectrum Its output ro(n) has a delay with respect to the artifact of biomedical signals, and analyzed the behaviour of caused by the movement of the thoracic cage. A delay the weights estimated by the LMS algorithm. In the of nd samples, larger than the delay in the path of the estimation of the spectrogram or even specific sensed respiration, is introduced in the path of the frequency content of signals, which is the case primary signal x(n). Number of taps in the adaptive investigated in our work, there are also other efficient filter is kept large enough to properly track the actual methods applied in the real world applications, such as delay. The output of the airflow sensor is multi component sinusoidal models [10,11]. There are synchronously related to the respiratory artifact, but it several quasi periodic biomedical signals which have is found to be deficient in frequencies above about 1 Hz interference of low frequency noises, such as and hence it is not effective in suppressing higher electrocardiogram (ECG), ZCG, brain evoked spectral components of the respiratory artifact which potentials and modulated signals in tele- severely affect the detection of the characteristic points communication applications. We made our tests in in the ICG waveform. A signal closely related to the impedance cardiography signals. The ZCG is an respiratory artifact can be estimated by using the output alternative method for helping the diagnostic of many of the respiration sensor and the sensed impedance heart diseases being simple, without trauma, non- signal together and used as the reference input for the invasive and has low effective cost. Through ZCG, LMS-based adaptive filtering. which is a quasi periodic signal, we can obtain the volume of blood being pumped by the heart, known as As the respiratory artifacts during inhale and exhale stroke volume (SV) as reported by [8,9,12–14]. phases are different, the reference needs to be estimated in synchronism with the respiratory phases. The two 2. SIGNALPROCESSING TECHNIQUE phases are detected from the sensed respiration ro(n), by taking the change in the polarity of its slope as the A schematic of the artefact suppression technique is onset of a new respiratory phase. The reference r(n) is shown in Fig. 2. estimated by a least-squares approximation based cubic B-spline [17],[18] fitted on xd(n), using a set of data points with equal number of uniformly spaced samples in each phase and the time indices corresponding to the two end points of each respiratory phase forming the knot vector. The reference input r(n) is filtered with the M-tap FIR filter, with coefficients wn(k), resulting in the output

(1)

The FIR filter output is subtracted from the

delayed input xd(n) to get the denoised output

(2) Fig.2. Adaptive filtering technique using estimated respiration reference. x(n) = sensed ICG signal (sum of the ICG s(n) and The output is used as the error e(n) for adaptive the artifact ra(n)), ro(n) = output of the respiration estimation of the filter coefficients by the LMS sensor, = output signal. algorithm. It uses an instantaneous estimate of the gradient vector, based on sample values of the tap input The output signals from a thoracic impedance sensor vector and the error for dynamic adaptation to adjust and a respiration sensor are simultaneously acquired filter coefficients on sample by sample basis [19], [20], and used as the inputs to the adaptive filtering block for using the equation suppressing the respiratory artifact. Signal x(n) is the sensed ICG. It is a sum of the desired signal s(n) and (3) the respiratory artefact ra(n), and is the primary input for adaptive filtering. The output of the respiration sensor ro(n), related to the respiratory artifact ra(n),is The step-size parameter µ satisfying the condition 0 < assumed to be uncorrelated to the signal s(n). In our µ< 2/M (mean-square (r(n)), is selected for controlling setup, a thermistor based airflow sensor, placed in front the convergence and stability of the adaptation. of the nostrils, senses the respiration.

174

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

3. ADAPTIVE LINEAR SYSTEM

In this paper we use the FLC and the LMS to estimate the Fourier coefficients of a given frequency band the signal, in this case, around the fundamental frequency, ѡs, of the acquired signal, dk. This method is known as LMS spectrum analyzer [15]. The reference inputs are pairs of sines and cosines at the designated frequency band of dk. The output of the adaptive LMS algorithm, yk, is subtracted from the acquired signal, which is corrupted by noises generating an error ek which is Fig.3. Block diagram of the FLC. sk is a minimized by the algorithm. The weights updated in deterministic signal or ZCG signal and nk is a low the process are Fourier the coefficients varying along frequency noise. time that we will call the LMS spectrogram. We use the LMS algorithm as a spectrum analyzer as it 3.1. Methods: was carried out by Widrow [15], in this work he discussed the relationship between the DFT and the A periodic of quasi periodic signal can be expressed as input signal, sk, and the vector weight, Wk, of the a Fourier expansion and therefore can be reconstructed algorithm. This algorithm calculates the possible as variations in time which result in an instantaneous output, yk, which is given by the internal product (4) between the reference inputs, Xk, and the vector weight, Wk. Mathematically this is presented as follows: where is a known frequency, Cik, are the (6) coefficients of the Fourier series. For periodic signals, the coefficients of the series are not time variant. where W is composed of updated weights: However, in the case of the quasi periodic signals, the k behaviour of the coefficients of the Fourier series is time variant, since the period of the signal cycles is not . constant. The adaptive linear system proposed in this work consists of an Adaptive linear combiner(ALC),developed by Vaz and presented in [6], This vector is initialized with the value zero, as used by characterized by the signal which we intend to analyze, Widrow et al. in [15]. Consider a noise, vk, inherent to dk, as presented in Fig. 3, and a group of reference input the acquisition process of the signal dk, which is vectors, Xk, defined as stationary and not correlated to such signal as in each component of the Fourier series. The method employed to adjust the weights of the adaptive system is the steepest descent method, as described by [3]. In (5) our work we used only some coefficients closer around the fundamental frequency of the acquired signal dk.

Where [0, ] corresponds 3.2. Weights behavior of the LMS algorithm to the frequencies of the desired signal in function of time and expressed in terms of the discrete time index In this section we demonstrate that low frequency k. H denotes the Hermitian operator. disturbances present in acquired signal interfere on the estimation of LMS spectrogram. We consider low noises as being those disturbances below the fundamental frequency of the signal. Let ѡs the fundamental frequency of the acquired signal and ѡv the frequency of the noise vk. Here, we consider that ѡs>> ѡv which causes a large frequency mismatch. We begin this analysis starting from the adaptation rule of the LMS algorithm [3]:

175

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

(7) Based on a trigonometric identity it is possible to

evaluate that the term Vk-1 results in two where µ is a real positive number which represents the sinusoidally time variant components around the size of the step or learning rate and controls the frequency of the correspondent reference input in system stability. Let us define the error , for each weight. This will affect the estimation and the signal of the weights, therefore we call it sinusoidal where sk is the desired signal and perturbation factor (SPF). vk is the additive low frequency noise. 4. RESULTS Substituting ԑ in Eq.(7), the updated weights become k The adaptive system was implemented as illustrated in Fig.3.In the ﬁrst experiment we used a sinusoidal (8) signal,sk with fundamental frequency of 1.2 and 0.01 Hz additive sinusoidal noise, vk. We used the LMS algorithm to obtain the spectrum of weights, Let us define a 2N × 2N diagonal matrix D with the LMS spectrogram, from the estimation of the diagonal elements, D(q,q)= , signal. The frequency band of the reference inputs was in the range 1.0–1.5 Hz, in steps of 0.05 Hz, q = -N,-(N-1),………-1,1,….N and giving a number of 11 harmonics. The step size parameter used was µ= 0:01.

. In Fig. 4a we show the LMS spectrogram of thetheoretical acquired sinusoidal signal sk adde

d to the noise nk. In order to show the existence of The new inputs will be expressed by the disturbance caused by the low frequency noise nk, in Fig. 4b is exhibited the LMS spectrogram and its of the signal dk with the noise eliminated. To ﬁlter the signal for the cancelation of n , we used a conjugate . k highpass, 4th order, butterworth digital ﬁlter with 0.1 Hz cutoff frequency. Applying the expectation operator to Eq. (8) and doing the appropriate considerations we obtain the following To validate this result, we also carried out some result: simulations with real biomedical signals.

-K -1 E[WK ]=2µDX0 E[D ] E[Vk-1](1- )(1- ) + W*(9) The ZCG signalwas acquired with noises such as: m ovement artifacts,breath signals, Electromyography where * is the optimal weight and is defined by signals (EMG)and electrocardiography signals (ECG), as well as other low frequency noises, taken from a patient at rest. In Fig. 5, we show in the LMS . spectrogram of the LMS algorithm obtained by the estimation of the measured ZCG signal and its filtered The equation shows the convergence of the system. version using the same filtering process of the previous simulation. We also carried out a simulation including The term contains the reference input sinusoids, a constant term or bias as proposed by Vaz [6] to therefore, its expected value tends to zero. So E[Wk] reduce the influence of the low frequency noise or bias. tends to the optimal weight. However, if we make the To evaluate the influence of the constant term we analysis of this equation considering as m all analysis computed the LMS spectrogram of the same ZCG window for the expected value or even without it we signal using the same parameters employed in the notice that the system is disturbed in the convergence simulation shown in Fig. 5, also including the constant term in the reference inputs. In Fig. 5, we show the by the term Vk-1. LMS spectrogram.

176

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The weight referent to the bias is not shown in the term. In (a) we show the spectrogram inﬂuenced by figure. In the last simulation, Fig. 6, we show the LMS the low frequency noises. Theseare due to thermal noi spectrogram of the same ZCG signal employed on the se, movement artifacts,breathing, etc. In (b) we show previous simulations using at the reference inputs the the LMS spectrogram of the highpass ﬁltered version frequency interval from 0 to 2 Hz, that is, we are also of the ZCG signal. estimating the weight components referent to the low frequency noise. This way there is no frequency mismatch caused by the low frequency noise.

Fig. . 6. LMS spectrogram of ZCG signal from a Fig. 4. LMS spectrogram for a theoretical simulate patient at rest. The LMS spectrogram was calculated d sinusoidal signal. The learning rate to estimation using a limited frequency band (0.5–2.0 Hz) including are µ=0:01. In (a) we show the LMS spectrogram a constant term. The weight referent to the a constant of a sinusoidal term is not shown. In (a) we show the spectrogram signal with frequency of 1.2 Hz with a low frequenc influenced by the low frequency noises. In (b) we show y additive sinusoidal noise of 0.01 Hz. Observe the the LMS spectrogram of the high-pass filtered version presence of ripple effect along the spectrum. of the ZCG signal. In (b) we show the spectrum of the highpass ﬁlter ed sinusoidal signal. It is possible to observe that t here is no ripple in the spectrum of the ﬁltered signal.

Fig. 6. LMS spectrogram of ZCG signal from a patient at rest. The LMS spectrogram was calculated using a

limited frequency band comprising the frequency band Fig. 5. LMS spectrogram of ZCG signal from a of the noise, that is, from 0 to 2.0 Hz. In (a) we show patient at rest. The LMS spectrogram was calcul the spectrogram that presents little influence due to low ated using a limited frequency band (0.5–2.0 Hz) frequency noises. In (b) we show the LMS spectrogram not Including a constant of the high-pass filtered version of the ZCG signal.

177

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

5. CONCLUSION [11] F. Gianfelici, G. Biagetti, P. Crippa, C. Turchetti, Multicomponent AM-FM representations: an Application of LMS-based adaptive filtering was asymptotically exact approach IEEE Transactions investigated for suppressing the respireatory artifact, on Audio, Speech, and Language Processing 15(3) by estimating a reference by spline fitting on the ICG (March 2007) 823–837. signal in synchronism with the respiratory phases [12] B. Santhanam, P. Maragos, Multicomponent AM- detected from the output of a respiration sensor. While FM demodulation via periodicity-based algebraic the spectra of the sensed respiration signal were separation and energy-based demo-dulation IEEE deficient in spectral component above about 1Hz, Transactions on Communications 48(3) (March spectra of the estimated reference closely approximated 2000) 473–490. the spectra of the respiratory artifact across the [13] T. Ishiguro, A. Umezu, Y. Yasuda, S. Horihata, subjects. Use of estimated reference in place of sensed A.K. Barros, Modified scaled Fourier linear respiration significantly reduced the filter tap length combiner in thoracic impedance cardiography and shortened the settling time to the duration of a Computers in Biology and Medicine 36(9) typical respiratory cycle. It was found to be particularly (September 2006) 997–1013. effective in suppressing the artifacts in recordings with [14] W.G. Kubicek, R.P. Patterson, J.N. Karnegis, D.A. a large variation in the respiration rate and cardiac Witsoe, R.H. Mattson, Development and activity. The technique does not require identification evaluation of an impedance cardiac of characteristic points and it is not affected by event output system Aerospace Medicine 37(12) variability. (December 1966) 1208–1212. [14] R.P. Patterson, W.G. Kubicek, E. Kinnen, D.A. Witsoe, G. Noren, Development of an electrical REFERENCES impedance plethys-mography system to monitor cardiac output in: Proceedings First Annual Rocky [1] V.K. Pandey, P.C. Pandey, Cancellation of Mountain Bioengineering Symposium, U.S. Air respiratory artifact in impedance cardiography in: Force Academy, Colorado Springs, USA, Proceedings of the 2005 IEEE Engineering in September 1964, pp. 56–71. Medicine and Biology 27th Annual Conference, 15. Widrow, P. Baudrenghien, M. Vetterli, P.F. Shanghai, China, September 2005, pp. 5503–5506. Titchener, Funda-mental relations between the [2] B. Widrow, S.D. Stearns, Adaptive Signal LMS algorithm and the DFT IEEE Transactions on Processing Series, first ed. Prentice-Hall New Circuits and Systems Cas-34(7) (July 1987) 814– Jersey, USA, 1985. 820. Evolutionary continuous genetic algorithm for [3] B. Widrow, M.E.T. Hoff Jr., Adaptive switching clinical decision support system circuits In: IRE WESCON Convention Record, vol. 4, New York, USA, June 1960, pp. 96–104. [4] S. Haykin, Adaptive filter theory Prentice-Hall Information and System Sciences Series, third ed. Thomas Kailath, Series Ed. New Jersey, USA, October 2005. [5] C.A. Vaz, N.V. Thakor, Adaptive Fourier estimation of time-varying evoked potentials IEEE Transactions Biomedical Engineering 36(4) (April 1989) 448–455. [6] C.A. Vaz, X. Kong, N.V. Thakor, An adaptive estimation of periodic signals using a fourier linear combiner IEEE Transactions on Signal Processing 42(1) (January 1994) 1–10. [7] Y. Xiao, Y. Tadokoro, Lms-based notch filter for the estimation of sinusoidal signals in noise Signal Processing 46(2) (October 1995) 223–231. [8] A.K. Barros, M. Yoshizawa, Y. Yasuda, Filtering non-correlated noise in impedance cardiography IEEE Transactions on Biomedical Engineering 42(3) (March 1995) 324–327. [9] A.K. Barros, N. Ohnishi, MSE behavior of biomedical event-related filters IEEE Transactions Biomedical Engineering 44(9) (September 1997) 848–855. [10]

178

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Stochastic Optimization Techniques as Effective Tools to Load Forecasting and Scheduling Using Distributed Energy Resources (DERs)

C.G. Monyei Department of Electrical and Electronic Engineering University of Ibadan [email protected]

ABSTRACT Curbing incessant and erratic power supply to halls of residence within the University of Ibadan Campus has been an impetus that has led to an upsurge in the number of proposals all geared towards providing solution to this critical problem. One of such proposals opines the design of a virtual power plant (VPP). In proposing such, the author seeks to address the problem on a double approach – tackle the erratic power supply and reduce carbon footprints. The proposal in achieving these aims takes advantage of the flexibility of Distributed Energy Resources (DERs), advancements in Information and Communications Technology (ICTs) and the stochastic nature of evolutionary algorithms (EAs) and artificial intelligence (AI) in creating a frame work for interaction between these components, the end users of electricity and the generation/distribution end. The crucial property of electricity being toyed with in the proposal is the ability of electricity to move in both directions depending on existing potential difference. A problem arising from this brilliant proposal though is the fact that loads have not been grouped or biased. The intermittent and stochastic nature of renewables limits their application to certain loads within the halls as such critical loads have to be connected to the school grid for uninterrupted supply. These loads could range from medical to cooking points. This paper seeks to address this issue of load biasing while taking advantage of stochastic optimization techniques in scheduling loads for supply and forecasting demand. The author in attempting to do this hopes to improve quality of supply and optimize demand among students within Independence Hall by suggesting creation of incentives, data mining to observe if a pattern exists which to a great extent mirrors students behavior and other EA tools which would prove useful.

Keywords: virtual power plant, evolutionary algorithms, artificial intelligence, distributed energy resources, stochastic . African Journal of Computing & ICT Reference Format C.G. Monyei (2013 Stochastic Optimization Techniques as Effective Tools to Load Forecasting and Scheduling Using Distributed Energy Resources (DERs). Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 179-184

1. INTRODUCTION

The University of Ibadan has not fared well in recent This sought to introduce some level of intelligence in times as regards power supply to its resident the corridor lighting system and also provide a case communities spanning the halls of residence, staff study from which relevant data could be extrapolated residential quarters, the Abadina Community etc. and used in designing a bigger VPP which could [12,13,14]. This poor power supply is no doubt gradually be introduced into halls’ power network and connected to poor power generation and an obsolete then gradually the school. While acknowledging the transmission/distribution network currently plaguing brilliance and importance of such a proposal, this Nigeria [9]. The University Management in tackling author attempts to identify some loopholes and design this menace has been exploring alternative means of flaws expected in fully implementing the proposal power generation like renewables, while also while also providing solutions to them. increasing the capacity of its diesel generators [9]. As can be surmised from Georgios et al [10], the These giant strides no doubt will yield no much change variable nature of generation of DERs is another as they do not seek to promote a culture of efficiency uncertainty that has to be addressed real time in and responsibility on the part of end-users as regards incorporating DERs into an existing grid. Their view is safe and efficient use of electricity. Monyei [9] in further reified by Sarvapali et al [4] who buttress the trying to correct this problem proposed the design of a fears associated with intermittent renewable energy virtual power plant for the corridor lighting system of sources while positing that smart grids must be able to Independence Hall as a pilot project. access these in supplying additional electricity. [1][2][3][5][6][7][8][11] also opine the earlier held notion on the intermittent nature of DERs.

179

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2. INTRODUCTION OF ADDITIONAL DESIGN CONCEPTS

From the foregoing therefore, the need arises for the introduction of additional design concepts to the earlier proposed model of Monyei [9]. In enhancing the earlier proposed design concept, the nature of loads available in the hall of residence is taken into account. This enhanced model calls for the discrimination of loads into different distribution centres for effective utilization of installed DG units and the University Power Supply Scheme.

Table 1 below gives an estimate of on-peak load and average off-peak load consumption for five days (Monday-Friday) in Independence Hall, University of Ibadan.

Table 1: School days load consumption estimate for Independence Hall Discriminated Time of the day to Peak consumption Average off-peak load come on (KW) consumption (KW) Lighting points 5pm – 7am 16.00 15.98 Monday - Friday Cooking points 24 hors 164.00 159.00 Corridor lighting 5pm – 7am 2.80 2.78 others 24 hours 45.00 41.00 Total (KW) 227.80 218.76

Table 2 below gives an estimate of on-peak load and average off-peak load consumption for weekends (Saturday and Sunday) in Independence Hall, University of Ibadan.

Table 2: Weekend load consumption estimate for Independence Hall Discriminated Time of the day to Peak consumption Average off-peak load come on (KW) consumption (KW) Lighting points 24 hours 16.50 16.25 Saturday Cooking points 24hours 170.00 168.00 Corridor lighting 5pm – 7am 2.80 2.78 others 24 hours 50 49.55 Total (KW) 239.30 236.58 Lighting points 24 hours 16.30 16.16 Sunday Cooking points 24 hours 165.00 163.00 Corridor lighting 5pm – 7am 2.80 2.78 Others 24 hours 47.00 45.00 Total (KW) 231.10 226.94

From the foregoing therefore, fig 1 below is proposed. 3. DESIGN ISSUES The proposed figure 1 gives a pictorial view of the Figure 2 shows the proposed connection of the loads to intended load discrimination. As can be observed, the available power sources – the grid and installed DG loads have been divided into the component structures Units. The connection being proposed is capable of analyzed in tables 1 and 2 above namely: lighting intentional islanding in the event of power failure from points, cooking points, corridor lighting and others. The the grid or a fault in the grid supply. As can be lighting points division represents all points of surmised from fig 2, the loads are connected in such a illumination within each room of residence within the way that the installed DG Units are capable of meeting hall. The cooking points refer to the access points in excess the demand from the lighting points and the provided by the University Management for students to corridor lighting point as shown below. access larger currents through the use of electric cookers and heaters. The corridor lighting (points) refer Let n be the total number of installed DG Units to illumination provided for hall ways, corridors, ………………………….…………… (1) kitchenettes, bathrooms and stairways. The designation n others refers to all other permitted points of accessing Let ∑Pdgi(t) be the total DG Capacity in Kilowatts electricity mainly within the rooms for lighter use. i=1 In coming up with this model, it is assumed that at time t ..………………………… (2) appropriate safety regulations and precautions such as fuse ratings, cable selection etc. have all been taken Let Plp(t) be the power demand in Kilowatts from into consideration. lighting points at time t………….... (3)

180

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Let Pcl(t) be the power demand in Kilowatts from intentional islanding. Incorporating this design concept corridor lighting points at time t…... (4) thus introduces some measure of intelligence and provides for extended intelligent supply in the event of a grid trip off. In analyzing figure 2, it is important to The connection is thus done such that note that switch 1 is only closed when grid supply is available and DG units are incapable of meeting current n n demands due to their capacity being either low, their ∑Pdgi (t) – Plp (t) – Pcl (t) ≥ (0.1 *∑ Pdgi (t)) being faulty or out of service (maintenance). Switch 2 i=1 i=1 is closed only when the DG Units are capable of ...... (5) meeting current demands and have their battery storage fully charged. The demand shifted from the other loads Equation 5 being very crucial allows for frequency – others and cooking points is analyzed subsequently. control and voltage regulation in the event of

To cooking To corridor lights Legend points

To corridor lights

To cooking points

To others To others To lighting points To lighting points Multi- distribution panel Fig 1: Load Discrimination

Let Pcp (t) be the power demand from the cooking points in kilowatts at time t .. (6)

Let Pop (t) be the power demand from other points in Kilowatts at time t ……… (7)

Let Pg (t) be the grid supply available at time t in Kilowatts ……………………. (8)

n n Pa (t) = ∑Pdgi (t) – Plp (t) – Pcl (t) ≥ (0.1 *∑ Pdgi (t)) …………………………... (9) i=1 i=1

Where Pa (t) is the available excess power from DG Units.

Load displaced by grid Pb (t) in Kilowatts is thus given as:

Pb (t) = Pcp (t) + Pop (t) - Pa (t) …………………………………….……….…. (10)

Power savings from the grid in Kilowatts is also computed as:

Pg (t) – Pb (t) ………………………………………………………………….. (11)

181

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Control signal From DG Units From grid Legend

To corridor lights

To cooking points

To others

2 To lighting points

Synchronizer

Intelligent

switch Fig 2: Proposed Load Connection to Grid and DG Units

Figures 3 and 4 shown subsequently are block diagrams showing the generation of control signal input into figure 2. As can be observed from figure 3, the voltage level as well as other parameters of the installed DG Units are checked against a reference regularly within a time frame t with the corresponding error signal generatedSum passed block on to an analyser which generates the relevant signal for transmission to figure 2 through wired or wireless connection. It is important to note here that the safety of the signals being transmitted be guaranteed+ and proven immune from hackers if wireless communication is to be used. Figure 4 on the other hand generates relevant signals which instruct the powerControl system to Vref Analyser either go into an islanding state or trip off connection from one supply source or both. The transmitted signalsignal is usually the difference between every available power supply source (grid and DG Units) at any time t and the load/demand at that same instant of time t. - VDG Fig 3: generating state of DG Units control signal

Sum block Sum block

+ + Control Control Pgrid Analyser Vref Analyser signal signal + - PDG Fig 4: generating control VDG Fig 3: generating state of - signal during excess load DG Units control signal Load demand Sum block

+ Control Pgrid Analyser signal

+ P Fig 4: generating control DG - signal during excess load Load demand

182

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

As can be observed from figures 1 – 4 above, REFERENCES incorporating these concepts not only shore up the reliability and safety of the earlier proposed model but [1] Alberto J. L., Tim M., Ray Z., Carlos E. M. and allows for more control of loads in terms of scheduling Lindsay, A. (2012) “Alternative Mechanisms for and displacement. In utilizing the varied benefits of Integrating Renewable Sources of Energy into artificial intelligence (AI) and other EA tools, the use Electricity Markets” retrieved from of Artificial Neural Networks (ANNs) and Genetic http://ieeexplore.proxy.library.carleton.ca/ Algorithm (GA) is thus proposed for the purposes of load forecasting and scheduling purposes. [2] Anderson L., Galloway S. and Stephen B. (2012) “Assessment of the Impact of different energy mixes in A single layer, five inputs, one output forward neural local decentralized energy networks” in Journal of network with no feedback is proposed for use in mining Power and Energy retrieved from the available data in a bid to forecasting subsequent http://pia.sagepub.com/ at Carleton University. load demands. The generated values are then checked against actual values and adjustment for errors is then [3] Pudjianto D., Ramsay C. and Strbac G. (2008) “ carried out on the ANN. This proposition is to be Micro-grids and Virtual Power plants: Concepts to incorporated into an earlier proposed work by Monyei support the integration of Distributed Energy [15] in which GA is used in optimally allocating Resources.” Retrieved from students into the halls of residence and calculating the http://ieeexplore.proxy.library.carleton.ca/ p.731. fees due each student in defraying the costs associated with increasing the capacity of the DG Units each year. [4] Sarvapali D., Perukrishnen V., Alex R. and Nicholas R. (2012) “Putting the ‘smarts’ into the smart 4. RESULT AND CONCLUSION grid: A Grand challenge for Artificial Intelligence” in Communication of the ACM, p.86 retrieved from As opined by Monyei [15], saving time and drudgery as http://ieeexplore.proxy.library.carleton.ca/ well as the flexibility of the GA in generating valid values make it the optimal tool in load forecasting and [5] Schafer A. and Moser A. (2012) “Dispatch scheduling. The combined use thus of the ANN and Optimization and Economic Evaluation of Distributed GA in effectively forecasting and scheduling load Generation in a Virtual Power Plant.” Retrieved from increases the flexibility, intelligence and extent of these http://ieeexplore.proxy.library.carleton.ca/ algorithms in arriving at a stable and optimal solution. This design is thus recommended for adoption in [6] Petersen M., Bendsten J. and Stoustrup J. (2012) designing a more improved virtual power plant for the “Optimal Dispatch Strategy for Agile Virtual Power halls of residence within the University of Ibadan. Plant” in the 2012 American Control Conference, Fairmont Queen Elizabeth, Montreal, Canada, retrieved from http://ieeexplore.proxy.library.carleton.ca/

[7] Such M. C. and Hill C. (2011) “Battery Energy Storage and Wind Energy Integrated into the Smart Grid” retrieved from http://ieeexplore.proxy.library.carleton.ca/

[8] Li D., Jayaweera S. and Abdallah C. T. (2012) “Uncertainty Modeling and Stochastic Control Design for Smart Grid with Distributed Renewables” retrieved from http://ieeexplore.proxy.library.carleton.ca/

[9] Monyei C. G. (2012) “Towards Sustainable Energy Development Using Virtual Power Plants” in the African Journal of Computing and ICT, retrieved from http://ajoict.net/ pp119-123.

[10] Chalkiadakis G., Robu V., Kota R., Rogers A. and Jennings N. R. “Cooperatives of Distributed Energy Resources for efficient Virtual Power Plants” Proc. of 10th Int. Conf. on Autonomous Agents and Multi-agent Systems – Innovative Applications Track (AAMAS

183

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

2011), Tumer, Yolum, Sonenberg and Stone (eds.), May, 2–6, 2011, Taipei, Tai-wan, pp. 787-794. [11] Kok J. K., Warmer C. J. and Kamphuis I. G. (2005) “Multiagent Control in the Electricity Infrastructure” _____ retrieved from http://ieeexplore.proxy.library.carleton.ca/ pp 75-82.

[12] http://www.channelstv.com/home/2012/04/26/Universit y-of-ibadan-spends-n40-million-on-power-v-c/

[13] http://www.ui.edu.ng/content/invitation-tender-o/

[14]http://Campusheathq.com/update-finally- university-of-ibadan-calls-off-strike-resumption-date- announced/

[15] Monyei C. G. (2012) “Adaptive Genetic Algorithm for Students’ Allocation to Halls of Residence Using Energy Consumption as Discriminant” (about to be published).

Author’s Brief

Monyei Chukwuka (B’ 1989) is a final year student in the Electrical and Electronic Engineering Department of the University of Ibadan, Ibadan, Nigeria. His interests centre on distributed generation as a means to reducing carbon emissions and improving energy efficiency and artificial intelligence in a bid to improving the flexibility and reliability of distributed energy resources. His final year project centers on modeling a virtual power plant for the corridor lighting system of a hall of residence within the University of Ibadan Campus. He has delivered a paper on this at the Nigerian Society of Engineers (NSE) Harmony Conference, 2012 and has several published works. He can be reached at [email protected]

184

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Introducing the Spatial Qualification Problem and Its Qualitative Model

P.C. Bassey *Computer Science Department University of Uyo UyoNiger, Nigeria [email protected]

B.O. Akinkunmi Computer Science Department University of Ibadan Ibadan, Nigeria [email protected]

ABSTRACT Spatial qualification problem is the impossibility of knowing an agent’s presence at a particular place at a certain time to be involved in an action or be participant in an event. The problem of spatially qualifying an intelligent agent requires commonsense reasoning which is qualitatively represented in qualitative spatial reasoning, a sub-field of knowledge representation and reasoning. In this paper, we present an overview of the spatial qualification problem and the qualitative formalism for reasoning with the problem. Existing spatial and temporal calculi for reasoning were combined and reused in the definition and axiomatization of basic concepts in the formalism. Quantified Modal Logic was seen to be suitable for the qualitative reasoning about these spatial concepts. The resulting spatial qualification logic (Alibi Logic) would be applicable in domains that require investigation of the problem of spatial qualification.

Keywords: Spatial qualification problem, Commonsense reasoning, Qualitative reasoning, Possible world semantics, Quantified modal logic . African Journal of Computing & ICT Reference Format P.C. Bassey & B.O. Akinkunmi (2013 Introducing the Spatial Qualification Problem and Its Qualitative Model Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 185-190

1. INTRODUCTION Formalisms where the investigation of spatial qualification in real life scenarios has been done In spatial domains, the problems of describing and through deep reasoning process made use of identifying an object, scene and route are common. probabilistic and fuzzy approaches. These approaches Attempts to solve these problems involve direct though it may lead to desired goals are quantitative. abstraction of the needed knowledge about the world Working with large sets of data is very expensive. and its properties such as actual size, weight and Also, these real world problems involve unproven ideas distance as seen in most formalisms. It also involves or assumed possibilities proposed for further the use of quantitative approaches and giving of precise investigation. Formulating most of these problems and predicted results. This approach is too expensive as using classical logics where only the truth value of a vagueness, uncertainty and granularity remains a formula is determined but not the way, mode and state problem in spatial and temporal domains. Spatial of the truth of a formula, will not lead to a logical knowledge apart from being vague and incomplete conclusion. (Galton, 2009; Cohn and Renz, 2008), is continuous, that is, it changes with respect to time. These Spatial qualification in our context is the possibility categories of problems involve commonsense that an agent could be present at a particular place at a knowledge (Cohn, 1999) which is best solved by certain time given prior antecedents. Reasoning about employing qualitative reasoning. A typical case of this spatial qualification involves an agent’s movement category of problems is that of investigating spatial from one region of space to another and the rate of qualification. change in time. Thus our research questions: Given prior antecedent to have been present at or absent from the scene of incidence under investigation, is it possible for the agent to have been at the scene of incidence at a certain time?

185

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

To have an answer to the above question, there is need 2. LITERATURE REVIEW to further ask: can spatial knowledge of this sort be Attempts to represent spatial knowledge have to do formally represented or is there a suitable language to handle incomplete/uncertain knowledge of this sort? with the several views about space. Representing space Can the representation be used in reasoning to reach as a concept, made researchers to start viewing space in the conclusion of “possibility” or “impossibility”? Can diverse ways. Casati and Varsi (Casati and Varzi, the validity of the formalism be proven? Formalizing 1997) presented two of the commonsense view of spatial qualification requires the use of a non-classical space: Newtonian and Leibniz view. Newtonian’s view logic such as modal logic due to the efficacy of its of space is that space is an individual entity in its own modalities in handling the “possible worlds” concepts. right independent of whatever entities may inhabit it. While Leibniz’s view contended with it by saying that The applicability of the spatial qualification logic is “there is no way of identifying a region of space except very promising in domains with high demand for by referencing what is or could be located or take place reasoning such as: at that region.” Reasoning with space requires categorization of the granularities of space and their o Alibi Reasoning: In a case where an accused relationship where several attempts to categorize person gives an alibi, to investigate the given ‘place’ as it relates with other spatial concepts as alibi to be true that there is no possibility of the neighbourhood, region, district, area and location have accused to be present at the scene of the been made (Bennett and Agarwal, 2007). incidence to be involved in the crime. The need to express location information about objects o Homeland Security: In a case of an ATM in space calls for simplifying the mathematical Fraud, the model if built into the ATM concepts by approximately referring to points without machines can help to investigate the possibility measure, that is, without employing the full power of of presence of an account holder at certain mathematical topology, geometry and analysis (Asher locations to carry out multiple transactions and Vieu, 1995). This approach is contrary to the that are spatially questionable due the time poverty conjecture by Forbus, Nielson and Faltings: difference between the repeated transactions. “there is no purely qualitative, general purpose o Planning: In planning, one needs to work out kinematics” (Forbus, 2008). They concluded by the feasibility of having an agent carry out an suspecting that the space of representations in higher action at some future time, given its current dimensions is sparse and for spatial reasoning, nothing location e.g. “I need to deliver a truck of less than numbers will do. oranges in Lagos in the next twenty minutes. I In an attempt to refute the poverty conjecture, increased am now in Ibadan which is about 2 hours from researches in Qualitative Spatial Reasoning (QSR) has Lagos.” addressed different concepts of space including The aim of this research is to formalize the logic of topology, orientation, shape, size and distance (Cohn, spatial qualification with respect to time using the 1999). Qualitative reasoning allows people to draw techniques of qualitative modeling (Forbus, 2008; useful conclusions about physical world without Cohn and Renz, 2008). Our formalism will provide a equations. It also allows one to work with far less data, logical framework for investigating the problem of than would be required when using traditional, purely spatial qualification. The achievement of the above quantitative methods. Frommberger (Frommberger, aim will result from the successful performance of 2008) pointed out that the use of this representation objectives such as: deciding an appropriate language empowers the agent to learn a certain goal-directed used for our logical theory and describing the axioms navigation strategy faster compared to metrical and derivation rules for our theory. The resulting representations, and also facilitates reusing structural formalized model (otherwise known as an alibi knowledge of the world at different locations within the reasoner) is deemed fit for the investigation of any same environment. Cohn pointed out that QSR is spatial qualification problem in the several domains. potentially useful, and that there may be many domains where QR alone is insufficient (Cohn, 1999). This The rest of the paper is organized as follows. Section 2 called for the addition of qualitative non-topological gives us an insight to the related literature. Section 3 information like orientation (Freksa, 1992), distance, discusses the methodology used in the formalization of size and shapes to the topological relations (Randell et the logic. The logic of spatial qualification is modeled al, 1992). The RCC-8 notations and their meaning are in section four with parameters used and the outcome as described in the table below. clearly represented using appropriate logical language in section 4. Section 5 gives the conclusion of the paper.

186

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Table 1. The RCC-8 Notations and Meanings Def5: EQ(l,l1)  P(l,l1)  P(l1,l)

S/No. Notation Meaning Def6: O(l,l1)  z(P(z,l)  P(z,l1)

1. EQ(l1,l2) l1 Equally connected with l2 Def7: PO(l,l1)  O(l,l1)  P(l,l1)  P(l1,l) 2. TPP(l1,l2) l1 is a tangential proper part of l2 Def8: EC(l,l1)  C(l,l1)  O(l,l1) 3. TPP (l2,l1) l2 is a tangential proper part of l1

Def9: PP(l,l1)  P(l,l1)  P(l1,l) 4. NTPP(l1,l2) l1 is not a tangential proper part

of l2 Def10: TPP(l,l1)  PP(l,l1)  z (EC(z, l)  EC(z,l1)) 5. NTPP(l2,l1) l2 is not a tangential proper part

of l1 Def11: NTPP(l,l1)  PP(l,l1)  z (EC(z, l)  EC(z,l1)) 6. DC(l1,l2) l1 has a disjoint connection with l2 The defined region connection relations are re-used in 7. EC(l ,l ) l is externally connected with l our logic to define the Regionally_part_of and the 1 2 2 1 Regionally_disjoint relations as follows. 8. PO(l1,l2) l1 is partially overlapping with l2 Def12: l,l1 Regionally_part_of(l,l1)  EQ(l,l1)  TPP(l,l1)  TPP(l1,l)  NTPP(l,l )  The effectiveness of these qualitative relations is fully 1 NTPP(l ,l) employed as models that have these combinations were 1 also created (Muller, 1998; Erwig et al, 1999; Bennett Def13: l,l1 Regionally_disjoint(l,l1)  et al; 2000). But these qualitative models are yet to DC(l,l1)  EC(l,l1)  PO(l,l1) address the qualification problem with respect to space. Attempts to address the spatial qualification problem Two representational languages are combined to obtain made use of probabilistic and fuzzy approaches. a suitable representational language that will help to Possible worlds here are arbitrary worlds of equally reason qualitatively about spatial concepts necessary divided grids of location with directional states of for investigating the spatial qualification problem. randomly assigned values (Dean et al., 1993) and set of Hence, the resulting representational language points with the reward function used to approximately combines First-Order logic because of its assign weights to the points (Shakarian et al., 2011). In expressiveness with the modal operators of Modal our work we view space as region rather than logic. considering geometric points and this allows reasoning Using a quantified (First-Order) modal logic (Fittings, without any randomly assigned value. 1998) leads to a new kind of semantic problem 3. METHODOLOGY however. The literature has a good number of papers trying to define a definite semantics for quantified A combined approach that employs both spatial and modal logic. One of the major problems in defining the temporal formalisms as its own formalism is adopted in semantics of quantified modal logics is the problem of our formal theory. The formalism makes use of certain having varying domains for different worlds within the existing spatial and temporal calculi for reasoning. framework of the possible world semantics. Because Some of these calculi have been stated and defined in the individuals of interest in our domain remain the the literature. Interestingly to us, the Region same, we are assuming that the objects in the domain Connection Calculus (RCC-8) and some temporal remain exactly the same as one move from one possible calculi which is either point based or interval based world to another. Consequently, in our logic, the were not left out. The definitions of the RCC-8 following Barcan’s axiom in (i) and (ii) hold: relations (Wolter and Zakharyaschev, 2000a, 2000b, 2002; Randell et al, 1992), which is based on the region x. P(x) x.P(x) connection relation, C, for the definition from the (i) literature of the eight disjoint pair of relations are as or follows: x.P(x)  x.P(x) Def1: l C(l,l)

Def2: l,l1 (C(l,l1)  C(l1,l)) (ii)

Def3: DC(l,l1)  C(l,l1)

Def4: P(l,l1)  z (C(x,l)  C(z,l1)

187

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Thus the structure of the Kripke model/Possible World road to traverse have to be known in other to determine Semantics (PWS) (Fittings, 2008) is best used to the minimum time it will take to traverse the road. For semantically explicate the model structure or the instance, if one wants to know how long it takes to formalism for our theory. A possible world is a traverse from Ibadan to Onitsha and he/she knows that universe in contrast with reality. It is also a region it takes an hour to traverse from Ibadan to Ijebu-Ode; 2 indexed with time. Kripke structure is defined by a hours from Ijebu-Ode to Ore; three hours from Ore to triplet, M = (W, R, V) where W is a non empty set of Benin; two hours from Benin to Onitsha, then it is possible worlds, R  W  W is the accessibility possible to say that it can take minimum of eight hours relation and V: (Prop  W)  (true, false) is the to traverse from Ibadan to Onitsha. valuation function. The meaning of the standard logical Our approach to solving this problem is based on operators: ,, ,  ,  and the quantifiers  and  qualitative modeling. Intelligent agents can use are as defined in the model semantics for First-Order qualitative models to reason about quantities without logic. The standard modal operators “necessity”  and having to resort to the nitty-gritty of mathematics and “possibility”  are as defined in the Kripke semantics. calculus. A particular approach that is powerful in this Something is necessarily true in our current world if it regard is that of discretization. The major determinants is true in all the worlds accessible from the current for our logic include the presence log, introduced be the world. Something is possibly true in the current world Present_at predicate and the accessibility of the if it is true in some world accessible from the current locations concerned, introduced by the Reachable world. predicate. The power of modalities of modal logic, Our formalization is based on the qualitative modeling necessarily and possibly, allows the representation of approach and the resulting system of axioms will be the uncertain spatial knowledge as shown in the axioms viewed in light of the Reified First-Order Predicate below. logic in which state propositions are treated as  x. l. t. Present_at(x,l,t)  (t1. t < t1 individuals and the standard modalities i.e. possibility Present_at(x,l,t )) (iii) and necessity are treated as properties of states. 1 Axiom (iii) gives the possibility of persistence of an agent. This states that for every agent x present at 4. THE QUALITATIVE MODEL OF SPATIAL location l at some time t, it implies that it is possibly QUALIFICATION true that the same agent is present at that location at a

later time t1. A qualitative reasoning model has been created to The reachability axiom that determines the possibility resolve the problem of spatial qualification. A of presence in our logic is as defined below: formalization of the solution to the problem of spatial qualification based on qualitative modeling has been x, l1, l2, t1, t2. made. Consider an agent that was present at place or Reachable(x, l1, l2, (t1, t2))  t1 < t2 location l and at a time t. Is it possible for the same  agent to be present at a difference place or location l1 at (Present_at(x, l1, t1)  a subsequent time t , given what was known? This 1 Present_at(x, l2, t2)) problem may be reduced in a sense to the problem of (iv) determining whether or not the agent can travel between one place or location l to another place or The following axiom gives the underlying idea of the location l1 between time t and time t1. A human reachability axiom. reasoning agent confronted with this problem would reason using the distance between location l and l1, and the speed or the rate at which the agent could travel.  x, l1, l2, t1, t2. Most human agents are able to estimate how long it takes to complete a journey on a certain highway (or Reachable(x, l1, l2, (t1, t2))  (t3, t4. t3 < t4  path). As can be affirmed by most people that this kind ((t4- t3) = (t2- t1))  Reachable(x, of reasoning is commonsense reasoning because it can l1, l2, (t3, t4))) (v) be answered experientially by anyone who has traversed the highway before or it can be estimated by anyone who knows the length of the highway. The person will use some prior knowledge of the distance and the speed limit allowed on the road. This knowledge can then be used to determine the time it will take simply by dividing the distance by the speed. It is obvious that the distance and the speed limit of the

188

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

In discretization quantities are divided into chunks. 5. CONCLUSION And the solutions to our problems can be deduced from The possibility of an agent to be present at a particular the solutions to the smaller versions of the problem. For example, if an agent being at location l at time t place at a certain time is viewed as a possible world in 1 1 our problem domain. This means that there is transition implies s/he can be in location l2 at a later time t2, and an agent being at location l at time t implies he can be between some or all the possible worlds in the set. 2 2 Some of these transitions may be possible while some at location l3 at a later time t3 and l3 is farther from l1 than l is, then x being present at l at time t implies x may not. Our interest in this problem is borne out of 2 1 1 the fact that the solution to this problem has many can be present at l3 at time t3. This is represented in axiom (vi) below: potential applications. Investigators in application domains like criminology, homeland security, planning,  x, l1, l2, l3, t, t2, t3. etc. will find our logical theory a useful companion Reachable(x, l1, l2, (t, t2))  required to reach possible conclusions about their Reachable(x, l2, l3, (t2, t3)) investigations. This serves as an abstraction mechanism  Reachable(x, l1, l3, (t, t3)). in an aspect of the formal ontology for the Semantic (vi) Web. Also, the absence of the agent can be inferred following Our logic treats any fact we know as something that the axiom below: remains permanently true. As such if we know that an agent is present at a location l at time t, then that fact is  x, l, l1, t. always true. We state in (viii) thus: (Present_at(x, l, t)  x, l, t. Present_at(x, l, t)   Present_at(x, Regionally_disjoint(l, l1)) l, t). (viii)  (Present_at(x, l1, t) This axiom represents the persistence of truth that for (vii) every agent x present at location l at time t, it implies The Regionally_disjoint used in axiom (vii) follows that it is necessarily true that every agent x is present at from the definition in Def13. location l at a certain time t. With the above system of axioms, our logical theory The above spatiotemporal logic answers the research should be able to make inferences that lead to the questions earlier mentioned. This formal theory would conclusion that it is possible or not possible for an be found wanting by companion systems in domains like criminology, homeland security against ATM object in a particular world, W1 at a certain time to fraud and planning. It’s usefulness in fraud detection in reach another world, W2. For any of these conclusions to be met, several logical axioms based on some stated ATM machines makes our logic a very useful logic that lemma and definitions about geographic space are will give a relaxed mind especially as we are imbibing required. The composition of definitions of the the cashless policy. This logic will also offer proofs of topological relations (RCC-8 relations) and the any given alibi such as the ones in forensic science to modalities with First-Order logic gives birth to our resolve legal issues. spatial qualification (alibi) logic. Quantified (First- Future work is on expressing and explicating the spatial Order) Modal logic for reasoning with this reasoning concepts in light of the Possible World Semantics problem is presented as a system of axioms. These (Kripke’s Model), analytically proving the logical logical axioms for inferring the possibility of an agent’s system for validity using tableau proof method and presence at a particular location at a certain time are integrating the logic into AI planning systems. For based on qualitative reasoning. instance some agents cannot perform certain actions except they are spatially qualified to do so. This calls for the need for a planning system to be able to reason Out of the scope of this paper is the formal semantics about spatial qualification. Further enhancement of the and the syntax of the logic presented to clarify the fact logic to reason about spatial qualification in a variable that our first-order modal logic is a fixed domain logic. world is also required. In other words the domain remains the same as one reasons from one possible world to another. With this system we argue that logic of presence such as ours satisfies all the properties of an S4 system of modal axioms which includes: K: (  )  (  ); T:    and 4:   .

189

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFERENCES

[1] Asher, N. and Vieu, L. (1995). Towards a [11] Frommberger, L. (2008). Learning to behave geometry of common Sense: A Semantics in space: a qualitative spatial Representation and a Complete Axiomatization of for robot navigation with reinforcement Mereotopology. In Proceedings of IJCAI’95. learning. In:International Journal on pp. 846-852. Artificial Intelligence Tools Vol. 17, No. 3. [2] Bennett, B. and Agarwal, P. (2007) Semantic pp. 465–482 © World Scientific Publishing Categories Underlying the Meaning of Place. Company. In Spatial Information Theory, 8th [12] Galton, A. P. (2009). Spatial and Temporal International Conference, COSIT 2007, Knowledge Representation. Earth Sci Melbourne, Australia, September 19-23, Inform. 2:169-187. Springer-Verlag. 2007, Proceedings (2007), vol. 4736 of [13] Muller, P. (1998b). A qualitative theory of Lecture Notes in Computer Science, motion based on spatiotemporal primitives. Springer, pp. 78-95. Principles of Knowledge Representation and [3] Bennett, B., Cohn, A. G., Torrini, P. and Reasoning. Hazarika, S. M. (2000). Region-Based [14] Randell, D. A., Cui, Z. and Cohn, A. G. Qualitative Geometry. Research Report (1992). A Spatial logic based on region and Series for School of Computer Studies, connection, Proceeding of 3rd International University of Leeds. Conference on Knowledge Representation [4] Casati, R. and Varzi, A. C. (1997). Spatial and Reasoning, Morgan Kaufmann, Los Entities. In Oliviero Stock (ed.), Spatial and Altos, pp. 55-66. Temporal Reasoning, Dordrecht: Kluwer, [15] Renz, J. and Nebel, B. (2007). Qualitative 1997, pp. 73-96. spatial reasoning using constraint calculi. In [5] Cohn, A. G. (1999). Qualitative Spatial Handbook of spatial logics, Springer. Representations. In: Proceeding of the [16] Shakarian, P., Dickerson, J. P. and International Joint Conference on Artificial Subrahmanian, V. S. (2011). Adversarial Intelligence IJCAI-99 Workshop on Adaptive Geospatial Abduction Problems. ACM Spatial Representations of Dynamic Tansactions on Intelligent Systems and Environment, pp.33-52. Technology, Vol. No. , 20. [6] Cohn, A. G. and Renz, J. (2008). Qualitative [17] Wolter, F., and Zakharyaschev, M. (2000a). Spatial Representation and Reasoning. In Spatial reasoning in RCC-8 with Boolean Handbook of Knowledge Representation. F. region terms. In Horn, W. (Ed.), Proceedings van Harmelen, V. Lifschitz and B. Porter of the 14th European Conference on (eds.). pp. 551-596. Artificial Intelligence (ECAI 2000), pp. 244– [7] Dean, T., Kaelbling, L. P., Kirman, J. and 248. IOS Press. Nicholson A. (1993). Planning with [18] Wolter, F., and Zakharyaschev, M. (2000b). Deadlines in Stochastic domains. In Spatio-temporal representation and reasoning Proceedings of AAAI93. based on RCC-8. In Cohn, A., Giunchiglia, http://lis.csail.mit.edu/pubs/pk/deanAAAI93. F., & Seltman, B. (Eds.), Proceedings of the pdf. 7th Conference on Principles of Knowledge [8] Fittings, M. (1998). On Quantified Modal Representation and Reasoning (KR2000), pp. Logic. 3–14. Morgan Kaufmann. http://comet.lehman.cuny.edu/fitting/boo [19] Wolter, F., and Zakharyaschev, M. (2002). kspapers/pdf/.../quantmodal.pdf. Qualitative spatio-temporal representation [9] Fittings, M. (2008). Possible World and reasoning: a computational perspective. Semantics for First Order LP. In Lakemeyer, G., & Nebel, B. (Eds.), http://comet.lehman.cuny.edu/fitting/book Exploring Artificial Intelligence in the New spapers/pdf/papers/FOLP.pdf Millenium, pp. 175–216. Morgan Kaufmann. [10] Forbus, K. D. (2008). Qualitative Modeling. In: Handbook of Knowledge Representation. F. van Harmelen, V. Lifschitz and B. Porter (eds.). pp. 361-393.

190

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

An Empirical Investigation of the Level of Adoption of Mobile Payment in Nigeria

Adebiyi Ayodele A., Alabi Esther, Ayo Charles K. and Adebiyi Marion O. Department of Computer and Information Sciences, Covenant University Ota, Nigeria. (ayo.adebiyi, esther.alabi, charles.ayo, marion.adebiyi{@covenantuniversity.edu.ng})

Corresponding Author: [email protected]@yahoo.com

ABSTRACT Mobile devices have been one of the most successful consumer products. In fact, Nigeria is the largest and fastest growing market in Africa. The telecommunications sector raked up about $8.42bn in revenue in 2008 and the number is expected to surge past $11bn by 2013. The proliferation of these Mobile phones and the need for a Cash-Less or Cash- Lite economy by the Nigerian Economy has therefore resulted in the introduction of Mobile payments in order to bridge the gap between the under banked and banking community. This study proposed a revised model that integrated Compatibility, Relative Advantage, Complexity, Trust and Security and Cost with Technology Acquisition Model (TAM) constructs (Perceived Usefulness, Perceived Ease of Use and Behavioral Intention to use) to investigate what influences Nigerian consumers’ adoption of Mobile Payment. In testing the model, a total of 250survey questionnaires were randomly administered to individuals in Lagos being the economic nerve center of the nation and transiting individuals from neighboring states but who are users of mobile phones. 227 questionnaires were received and used for further analysis. The results show that Nigerians appreciate the benefits of the introduction of the Cashless Economy via Mobile payment and they would also be encouraged to use it because of its benefits such as Convenience, Ease of Use, Ease of Access, Reduced time of transaction. However, the complexity of the interface and procedures, trust in the service provider and agents (vendors), security and privacy of valid information and cost are pertinent factors that affect the adoption and a successful implementation of Mobile Payment.

Keywords: Mobile Payment, Cashless, Economy, Mobile Commerce, TAM Model, IDT Model . African Journal of Computing & ICT Reference Format Adebiyi Ayodele A., Alabi Esther, Ayo Charles K. and Adebiyi Marion O. (2013 An Empirical Investigation of the Level of Adoption of Mobile Payment in Nigeria Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 191-202

1. INTRODUCTION

Information technology (IT) is fast becoming the most important factor in the future development of banking, When the Nigerian telecommunications sector was influencing banks’ marketing and business strategies. In deregulated in 2001, allowing the country to join the rest recent years, the adoption of e-banking began to occur of the world in acquiring the Global System for Mobile quite extensively as a channel of distribution for financial Communications, popularly known as GSM, with about services due to rapid advances in IT and intensive 422,000 subscribers nobody envisaged its far-reaching competitive banking markets (Mahdi and Mehrdad, 2010; benefits to the economy and people of Nigeria. Now, Dube, et.al., 2009). Electronic Commerce, (e-commerce) Mobile Payment or Mobile money, a new electronic and e-banking have since become a way of life in Nigeria payment scheme, is opening up a new vista of such that Nigerians purchase and sell goods and services opportunities, extending the multiplier effects of over the Internet. According to a customer survey report in telecommunications in the country. Mobile devices can be 2008, many Nigerians have adopted the service and the used in a variety of payment scenarios, such as payment transactions made accounted to N360 billion but it has not for digital content (e.g. ring tones, logos, news, music, or been able to address the on-the-go access due to the games), flight tickets and bus fares and also to pay bills barrier of a Personal Computer (PC) and Wireless and invoices but the payment of digital content has not technologies . been fully tapped into in Nigeria.

191

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Mobile payments are payments for goods, services, and Mobile Banking (M-banking) is used to denote banking bills with a Mobile device (such as a Mobile phone, smart- services and facilities offered by financial institutions such phone, or personal digital assistant (PDA)) by taking as account-based savings, payment transactions and other advantage of wireless and other communication products by use of an electronic Mobile device. Mobile technologies such as Mobile telecommunications networks Payment refers to the various components required to and proximity technologies. It is a phone-based cash deliver Mobile payment to the banking and non-banking savings and transfer system, which turns a GSM phone community (CBN, 2011). In Nigeria, Mobile Payment will into a savings account allowing the owner save money in help to curb the problem of long queues in banks; it is also it and from which he could withdraw or transfer money at very convenient as the users can have access to financial a later date. services at anytime. Economically, Mobile payment tends to increase the number of jobs in the country hence The Central Bank of Nigeria (CBN) led by Governor reducing the rate of unemployment. SanusiLamidoSanusi has recently passed a policy to take effect in 2012 requiring that all cash withdrawals and 2.1 Selected Global Payment Systems deposits be set at a daily limit of a maximum of N150, 000 In the Phillipines, Globe Telecom introduced “G-cash” in while pegging that of corporate entities at N1, 000,000, October 2004, an electronic money transfer facility that with penalty fees of N100 per extra N1,000 and N200 per turns a Mobile phone into an electronic wallet. It provides N1,000 imposed on individual and corporate defaulters services where G-Cash subscribers purchase goods and respectively. This policy is geared towards achieving a services over the counter or remotely, receive domestic Cashless Economy where all Nigerians (from children to and international remittances. It is a very convenient way the aged) can elevate from the traditional banking in the of carrying out transactions which also allows for payment cities; taking them to the streets where Mobile money of utility bills and Mobile phone airtime credits. agents are on hand, ready to do the same thing a customer normally does in the banking hall. In order to achieve this POCit is a Mobile payment solution designed by South policy 16 operators were recently licensed to provide the Africans; it has the widest reach of the other 12 Mobile Mobile payment service: 10 non-bank-led (Pagatech, payment solutions available in South Africa. It allows Paycom, M-Kudi, Chams, Eartholeum, E-Tranzact, customers send and receive money from any bank faster, Parkway; Monitise, FET, and Corporeti) and 6 bank-led all the payer needs is the cell phone number of the payee. (Stanbic IBTC Bank, Ecobank Nigeria, Fortis MFB, It is very unique because it does not require special SIM UBA/Afripay, GTBank/MTN, and First Bank of Nigeria). cards.

The primary objective of this research is to investigate the In Kenya, M-PESA (M for Mobile and PESA for Money level of adoption of Mobile Payment in Nigeria; analyzing in Swahili) is widely recognized as one of the most Nigeria’s complex environment and examine consumer’s successfully implemented Mobile payment service. It is a willingness to use a Mobile phone as an instrument for SMS‐based money transfer system that allows individuals initiating and conducting secure financial transactions. to deposit, send, and withdraw funds using their cell The rest of this paper is arranged as follows: Section two phone. M‐PESA has grown rapidly, reaching is the literature review, section three describes the research approximately 65 percent of Kenyan households by the model and hypothesis, section four presents the analysis end of 2009, and is widely viewed as a success story to be and results and section five the Conclusion and emulated across the developing world. Registration is recommendation of the study. simple; it requires an official form of identification (typically the national ID card held by all Kenyans, or a 2. LITERATURE REVIEW passport) but no other validation documents that are typically necessary when a bank account is opened. Due to the high rise in demand for conducting payment and other financial transactions everywhere and at anytime in Nigeria, Mobile payment has been suggested by the Central Bank of Nigeria as the solution to facilitate such micropayments in the country. Mobile Payments are defined as the use of a Mobile device to conduct a payment transaction in which money or funds are transferred from a payer to a receiver via an intermediary, or directly without an intermediary (Mallat, 2007). While this definition includes Mobile payment transactions conducted via Mobile banking systems, a distinction between Mobile payment and Mobile banking services should be noted. Mobile banking services are based on banks’ own legacy systems and offered for the banks’ own customers.

192

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

In Nigeria, there are three ways in which Mobile payment “PocketMoni” is a Mobile money service designed by is carried out: Etranzact International Plc; one of the licensed Mobile  Card Account Based money service providers to allow users conduct financial  Bank Account Based transactions anytime, anywhere from a Mobile phone. It  Stored Value (e-money) Account Based enables users pay DSTV, HiTV, MyTV and PHCN bills, buy airline tickets and book hotels. This service is flexible In the Card Based scenario, a payment card (Credit, Debit as users can send money to PocketMoni subscribers, an and Pre-paid) is linked to the Mobile phone for initiating Etranzact card, a bank account and any Mobile phone and concluding the Mobile payment financial transactions. user. The Bank Account Based is where the financial transactions are initiated through the bank accounts of the This service is available to users in two ways; a menu consumers which could be their existing bank accounts driven software that is downloaded and installed on a (current account, savings account, domiciliary account Java- enabled phone and using the SMS/IVR channel. The etc) in the various banks or newly generated ones. The SMS/IVR channel enables customers using a non Java- Stored Value Account Based is a scenario where financial enabled phone access the service. A service fee of N100 is transactions are driven through a system-based account. charged at agent locations and partner banks. Its service is Examples of the stored value are Re-loadable stored value secure, cost effective, convenient and available for accounts, prepaid accounts etc. everyone.

There are also three Mobile payment models for the 3. RESEARCH MODEL AND HYPOTHESIS implementation of Mobile payment services namely:

 Bank-Focused Model 3.1 Technology Acceptance Model  Bank-Led Model The Technology Acceptance Model is one of the models  Non-Bank Led Model that have been developed to provide a better understanding of the usage and adoption of Information Bank-Focused Model is a model which has a Financial technology. It is presently a prominent theory used in Institution as the Lead Initiator (an entity or representative modeling technology acceptance and adoption in of other partners), who is responsible for ensuring that the Information systems research. Fred Davis in 1985 various solutions and services within a Mobile payment proposed the Technology Acceptance Model (TAM) in his system meet the regulatory requirement of the Central doctoral thesis at the MIT Sloan School of Management Bank of Nigeria. The Bank-Led Model is a model where a (Davis, 1985). bank or other consortium of banks partner with other organizations to deliver banking services by leveraging on Two cognitive beliefs are posited by TAM: Perceived the Mobile banking system. It involves where there is usefulness and Perceived ease of use. According to TAM, collaboration between a licensed deposit-taking financial- one’s actual use of a technology system is influenced institution (deposit-money banks, microfinance banks and directly or indirectly by the user’s Behavioral Intentions, discount houses) and an organization verified by the Attitude, Perceived usefulness of the system, and partner banks. Perceived ease of the system. Perceived usefulness (PU) is defined as the degree to which a person believes that using 2.2 Mobile Payment Services in Nigeria a particular system would enhance his or her job UBA’s (United Bank of Africa), Nigeria’s foremost performance and Perceived ease of use (PEOU) is defined financial institution is one of the recently licensed as the degree to which a person believes that using a operators to provide the Mobile Payment service in particular system would be free of effort. Nigeria. It kicked off the first Nigeria’s Mobile Money service (U-MO) in collaboration with its associate In a recent empirical study conducted on Factors company Afripay Limited on November 24, 2011. It Affecting Malaysian Mobile Banking Adoption, Cheahet. enables phone users buy airtime for self and send to al., (2011) identified that Perceived Usefulness has a others, pay utility bills, pay for goods and services in positive relationship in examining the Intention to adopt shops and on the Internet as well. It is available to existing Mobile banking; which implies that if Mobile banking is UBA (United Bank for Africa) customers and non-UBA useful and beneficial, users are more likely to adopt customers, existing customers will find the service of great Mobile banking services. Similarly, Perceived Ease of use as they will be able to easily load money from their Use was found to have a positive relationship with bank accounts into U-MO and move money received on adoption of Mobile banking. In order to understand the U-MO account into their bank accounts. Customers Factors affecting the Intention to use, an Online Learning register online, at a U-MO agent location or at a UBA community argued that Perceived Ease of Use has a branch. After registration, they receive a confirmation signiﬁcant positive effect on Perceived Usefulness, and it SMS. Then U-MO Java application will have to be is also the determinant with the strongest direct impact on downloaded and installed on their mobile phones. Intention to Use.

193

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The impact of Perceived Ease of Use has on Intention to Complexity: It refers to the degree to which an innovation Use is not as strong as that of Perceived usefulness and is perceived as being complicated or easy to use. Previous Online Learning Experience (Liu et. al., 2009). It implies that when the system is easy to use, users feel it is Observability: It is watching an innovation and how it more useful; therefore, they will have stronger intentions works to know if it is safe and beneficial for use. It is the to use the Online learning community. In Healthcare degree to which the results of an innovation are observable related research, Wu et. al., (2007) showed that Mobile to others. Healthcare System (MHS) self-efficacy and Perceived Ease of Use have very strong total effects on the Trialability: It refers to the degree to which an behavioral intention in contrast, Perceived Usefulness innovation may be sufficiently tested prior to adoption. moderately affect the behavioral intention. From above, This is the degree to which an innovation can be the Technology Acceptance Model has been tested and experimented with on a limited basis. proven to enhance the behavioral intention to use Mobile payments. Hence, the following hypotheses have been Although much research supports the TAM as an excellent proposed: model to explain the acceptance of IS/IT, it is questionable whether the model can be applied to analyze every H1: Perceived Usefulness has a direct positive instance of IS/IT adoption and implementation. Many influence on the Behavioral Intention to use empirical studies recommend integrating TAM with other Mobile payment. theories (e.g. IDT, or DeLone& McLean’s IS success model) to cope with rapid changes in IS/IT, and improve H2a: Perceived Ease of Use has a direct positive specificity and explanatory power (Carter & Belanger, influence on Perceived Usefulness of Mobile 2005; Legris et. al., 2003). Technology Acceptance Model payment. and Innovation Diffusion Theory are extremely similar in some constructs and also supplement each other. While H2b: Perceived Ease of Use has a direct positive Relative advantage is similar to Perceived Usefulness, influence on the Behavioral Intention to use Complexity is similar to Perceived Ease of Use; an Mobile payment. integration of the two would provide a stronger model (Wu et. al., 2007). 3.2 Diffusion of Innovation Theory This theory was developed by Rogers (1983). He The work of Lee et. al., 2011, that combined IDTwith explained the process of Innovation diffusion as one TAM in Employees’ Intentions to use E-Learning which is dictated by uncertainty reduction behavior Systems, showed that Compatibility and Relative amongst potential adopters during the introduction of advantages has significant positive effects on Perceived technological innovations. Innovation Diffusion Theory Usefulness; this implies that before the employees in an (IDT) consists of six major components: innovation organization can adopt the e-learning systems, they need characteristics, individual user characteristics, adopter to be convinced that it meets their job needs or is relevant distribution over time, diffusion networks, innovativeness to their job and be assured that it would be useful to them. and adopter categories, and the individual adoption Rokhman, (2011) used Innovation Diffusion Theory to process. Arguably the most popular of the six components study E-government adoption in developing countries. The of IDT centers on the characteristics of the innovation findings revealed that the degree of Relative advantage itself. After analyzing a variety of previous innovation can be used to predict Internet users’ probability to adopt diffusion studies, Rogers (1983) singled out the following E-government; Intention to use E-government services five characteristics of innovations that consistently increases as Relative advantage increases, the users’ influence the adoption of new technologies (Green, 2005): would find it useful as it would enhance their efficiency and make it easier in interacting with government Relative Advantage: This is the degree to which an agencies, Intention to use E-government also increases as innovation is perceived as better than the idea it a Compatibility increases. It was also found that image supersedes by a particular group of users, measured in and Ease of use are not good predictors of intention to use terms that matter to those users, like economic advantage, E-government. social prestige, convenience, or satisfaction (Robinson, 2009).

Compatibility: This is the degree to which an innovation is perceived as being consistent with the values, past experiences, and needs of potential adopters. An idea that is incompatible with their values, norms or practices will not be adopted as rapidly as an innovation that is From the literature review above showing that Innovation compatible (Robinson, 2009). diffusion theory can be used to carry out a valid research, the following hypotheses can be proposed:

194

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

H3a: Compatibility has a direct positive influence on H6a: Trust and security has a direct positive influence Perceived Usefulness of Mobile payment. on Perceived Usefulness of Mobile payments.

H3b: Compatibility has a direct positive influence on H6b: Perceived Ease of Use has a direct positive Perceived Ease of Use Mobile payment. influence on Trust and security of Mobile H4: Relative Advantage has a direct positive payments. influence on Behavioral intention to Use Mobile payment. H6c: Trust and security has a direct positive influence on Behavioral intention to Use Mobile H5: Complexity has a direct positive influence on payments. Behavioral intention to Use Mobile payment. 3.4 Cost of Service In the review of previous researches on Mobile payment 3.3 Trust and Security adoption and acceptance, the cost of the service is an Trust and security is undoubtedly one of the most essential factor to its adoption. Dahlberg et. al., 2008, significant factors that influence consumer adoption of a identified important adoption factors for Mobile payment new innovation that involves electronic transactions. services to be Ease of use, Trust and security, Usefulness, Being able to convince Nigerian Consumers that Mobile Cost, and Compatibility. Mallat, (2007) also identified it payment will be free from fraud and would protect their as an important factor that should be examined separately. privacy would be a critical success factor. Perceived Setting up a Mobile payment service that is transparent in security and trust in vendors and payment systems is a the pricing and does not cost so much would improve the significant determinant of Mobile commerce success level of adoption. The costs may include direct transaction (Mallat, 2007). Shneiderman (2000) argues that improving costs, fixed costs of usage and the cost of the technical positive security and privacy perceptions are most infrastructure for the customer (Khodawandi et al. 2003). important for sustained activity in electronic commerce Nigerians will only want to use a service from a trusted and more importantly Mobile payments. Mobile operator or bank who will maintain a stable price over a period of time. Dahlberg et. al., (2008) proposed the Trust enhanced Technology Acceptance Model. The theoretical basis of In order to test for validity, the following hypotheses have their study premise on the TAM model to measure been proposed: whether it provides comprehensive explanation for consumer decisions related to adoption of Mobile H7a: Cost has a direct positive influence on Perceived payments.The need for trustworthiness is not limited to Usefulness of Mobile payments. Mobile payment service provider but includes merchants as well. Consumers would only want to conduct business H7b: Cost has a direct positive influence on Perceived with merchants who are established and are Ease of Use of Mobile payments. trustworthy.The construct Trust and security has been tested and proven to enhance the prediction of behavioral H7c: Cost has a direct positive influence on Behavioral intention to use Mobile payments. Hence, it has been intention to Use Mobile payments. incorporated to enhance this study and the following hypotheses have been proposed:

195

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

H7c Cost Compatibility H3a

H7a Perceived H3b H7b bbbb Usefulness H1 H6a ba

Trust and H2a

Security Perceived H2b Behavioral Intention to H6b Ease of Use Adopt Mobile H4 Payment

Relative H5 Advantage

Complexity

H6c

Figure 1: The Research Model

4. DATA COLLECTION INSTRUMENT The questionnaires were administered by personally approaching businessmen and women, students, at The Data collection Instrument used in this study is a markets, malls, on the streets banks, at offices to survey questionnaire with 38 questions. In order to ensure participate in the research. The questionnaires were also content validity the questions were formulated from other translated to Yoruba and Nigerian Pidgin English for the research studies (Rogers, 1983; Chau& Hu, 2001; Davis unlearned. Transiting business men and women who are et. al., 1989, Taylor & Todd, 1995). The items in the also non-residents of Lagos from areas such as Abuja, Epe survey instrument were developed by adapting existing and Ota, Ogun State were participants in this research. measures validated by other researchers in Mobile banking and Mobile payment environment, or by converting the 4.3 Analysis of Data and Presentation of Results definitions of the construct into a questionnaire format. The participants in this study were consumers’ involving Each participant was asked to indicate his or her degree of young and old adults, business men and women, bankers, agreement using the 5 point likert scales: Strongly Agree, casual workers, academics, traders, students who were Agree, Disagree, Strongly Disagree and Neither Agree nor randomly selected. A total of 250 (Two hundred and fifty) Disagree. The obtained data were analyzed based on the questionnaires were distributed, 227 (Two hundred and correlation and regression analyses using SPSS (Statistical twenty seven) were received giving a response rate of package for social sciences) version 19. 90.8%, 14 (fourteen) were dropped due to missing data or invalid responses resulting in 5.6%, 9 (nine) were not returned which is 3.6% of the research population. 4.1 Study Population Since Mobile Payment/ Mobile Money is being tested in 4.4 Demographic Variables and Mobile Phone and Lagos, Nigeria and the research study is geographically Banking Usage limited to Lagos, data was collected in few commercial 197 respondents from Lagos; making 86.8% of the total centers such as Yaba, Victoria Island, Lekki, Ejibgo, Ikeja, data, 13 were from Ota, Ogun State which makes 5.7%, 6 and Lagos Island etc. from Epe; 2.6%, 1 from Abuja; .4% and 10 from other neighboring states in Nigeria making the remaining 4.4%,

196

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

there were slightly more male respondents than females. 227 respondents (98.2%) have a Mobile phone while the 117 (51.5%) were males and 110 (48.5%) were females, remaining 4 people (1.8%) do not possess a Mobile phone. 34.4% were between 20 and 29 years of age (young 2.2% claimed to have been using the Mobile phones for adults) which had the highest response rate. 28.2% of the less than a year, 37.0% between 1-5 years, 44.9% between respondents have a minimum of high school certificate, 6-10 years, 10.1% have been using the Mobile phone for 47.6% were either undergraduates or have a minimum of a more than 10 years, 4.0% were not aware of how long BSc. Degree, 22.0% were masters’ degree holders and they have had the Mobile phone. The missing system 2.2% have a PhD.The results also showed that a fair indicates that there was no entry for 4 respondents in this 38.3% of the respondents were employed. On the Mobile question. The Table below provides a detailed phone ownership, the results show that a large number of demographic profile. the respondents have and use a Mobile phone. 223 out of

Table 1: Demographic of the Respondents Frequency Valid Percent Location

Lagos 197 86.8 Ota,Ogun State 13 5.7 Epe 6 2.6 Abuja 1 .4 Others 10 4.4 Gender Male 117 51.5 Female 110 48.5 Age group Less than 20 64 28.2 20-29 78 34.4 30-39 60 26.4 40-49 20 8.8 More than 49 5 2.2 Education High School Certificate 64 28.2 Undergraduate 108 47.6 Masters Degree 50 22.0 PhD 5 2.2 Employment Employed 87 38.3 Self Employed 32 14.1 Not-Employed 5 2.2 Student 98 43.2 Employed and student 5 2.2 Mobile Phone Ownership Yes 223 98.2 No 4 1.8 Duration Less than a year 5 2.2 1-5 years 84 37.0 6-10 years 102 44.9 More than 10 years 23 10.1 Unknown 9 4.0

197

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Respondents were asked to indicate the rate at which they use any of the various form of Internet Banking and Mobile Banking system. Table 2 below provides adetailed frequency distribution of the respondents’ e-banking and m-banking usage. 81.9% of the respondents were reported to belong to the banked populace (have bank accounts) while the 18.1% are the unbanked populace (have no bank accounts). 66.5% of the participants have never made use of the Internet banking services, 2.6% use the service weekly, 22.0% use it less than monthly and 8.8% make use of the service at least once in a month.

Table 2: Respondents’ usage e-banking and m-banking services Frequency Valid Percent Bank Account Ownership Yes 186 81.9 No 41 18.1 Internet Banking Use Never 151 66.5 Less than Monthly 50 22.0 At least once in a Month 20 8.8 Weekly 6 2.6 Mobile Banking Use Never 185 81.5 less than Monthly 22 9.7 At least once a Month 15 6.6 Weekly 5 2.2 Mobile Payment Awareness Yes 179 78.9 No 48 21.1 Activity Performed Mobile Money transfer 11 4.8 Airtime transfer 11 4.8 Payment of bills and other services 1 .4 Mobile banking 11 4.8 Mobile Money and Mobile banking 4 1.8

198

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4.5 Reliability Analysis Reliability of the constructs was estimated using the Cronbach’s alpha (α) to determine the internal consistency of the measurement items in the construct; it is the extent to which all of the items measure the same variable. As shown in the table 3 below, the Cronbach’s alpha (α) values range between 0.584 and 0.902. For each of the constructs, they were above the 0.70 benchmark recommended by Bagozzi& Yi (1988) except Complexity with a value of 0.584.

Table 3: Reliability analysis of the constructs Determinants Number of Items Cronbach’s alpha PU 3 0.713 PEOU 4 0.839 BI 3 0.731 RELATIVE ADVANTAGE 4 0.833 COMPATIBILITY 3 0.902 COMPLEXITY 2 0.584 TRUST AND SECURITY 3 0.683 COST 3 0.686

4.6 Correlation Analysis Pearson’s correlation analysis was carried out to explore the strength and direction of the relationship between the variables (dependent and independent) for each hypothesis. The results show all positive values. The Table 4 below shows that there is a positive relationship between Behavioral Intention to Use Mobile payment and its variables; Relative Advantage, Compatibility, Complexity, Trust and Cost.

199

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Table 4: Correlations Analysis of the constructs PEOU PU BI RA Compatibility Complexity Trust Cost PEOU Pearson 1

Correlation Sig. (2-tailed) N 227 PU Pearson .433** 1

Correlation Sig. (2-tailed) .000 N 227 227 BI Pearson .456** .364** 1

Correlation Sig. (2-tailed) .000 .000 N 227 227 227 RA Pearson .540** .402** .475** 1

Correlation Sig. (2-tailed) .000 .000 .000 N 227 227 227 227 COMPATIBILTY Pearson .384** .349** .347** .335** 1

Correlation Sig. (2-tailed) .000 .000 .000 .000 N 227 227 227 227 227 COMPLEXITY Pearson .278** .269** .245** .375** .316** 1

Correlation Sig. (2-tailed) .000 .000 .000 .000 .000 N 227 227 227 227 227 227 TRUST Pearson .306** .317** .305** .300** .174** .447** 1 Correlation Sig. (2-tailed) .000 .000 .000 .000 .009 .000 N 227 227 227 227 227 227 227 COST Pearson .443** .351** .418** .493** .395** .353** .477** 1 Correlation Sig. (2-tailed) .000 .000 .000 .000 .000 .000 .000 N 227 227 227 227 227 227 227 227 **. Correlation is significant at the 0.01 level (2-tailed).

200

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

4.7 Hypothesis Testing 5. CONCLUSION Multiple Regression analysis is the technique employed to address how well a set of variables is able to predict From the research conducted, it was noted that the level an outcome and which variable in the set of variables of adoption of consumers in Nigeria is promising and gives the best outcome in this study. It is the technique this can be linked to the several advantages that is that was employed to test the hypothesis. In the first associated with the use of Mobile payment such as ease Multiple Regression test, the Beta (β) value with the of use, ease of access, reduced time of transaction etc. largest coefficient is .234 (Relative Advantage); it is Nigerians also appreciate the benefits of the the variable that makes the strongest unique introduction of the Cashless Economy via Mobile contribution in explaining the Behavioral Intention to payment but there are still some factors that can hinder Use. Perceived Usefulness (β =.113, P <0.01), its adoption by the Nigerian populace. The Mobile Perceived Ease of Use (β = .197, P< 0.01), Relative operators need to be aware that although there is Advantage (β= .234, P< 0.0005) and Cost (β=.144, P< inadequate infrastructure to cover the entire country at 0.05) made a unique and statistically significant kick-off, there is also the need to establish relationships contribution to the prediction of the Behavioral with their prospective users at the pilot phase; they Intention to Use Mobile payment. need to enroll trustworthy agents that are handling the money.

With Sig = .267(P > 0.1) and Sig= .845 (P > 0.1) Trust Relative Advantage was found to be the strongest and Security and Complexity respectively do not make significant determinant of Intention to Use (β=.234) a significant unique contribution to the prediction of and this supported the studies of Pikkarainenet. al., Behavioral Intention to use Mobile payment. Hence, (2004) Venkatesh&Davis (2000) and Cheahet. al., H1, H2b, H4 and H7c are supported while H5 and H6c (2011). Due to the ease of mobility of a Mobile phone were not supported.In order to determine the impact of and the need for a convenient way to perform financial the following variables such as Trust, Com patibility, activities, the relative advantage was considered the Perceived Ease of Use, Cost on Perceived Usefulness, most significant factor. For example in the areas where another test was conducted. Perceived Ease of Use (β people have to travel some distance to have access to = .278, P< 0.0005) Trust (β= .163, P< 0.05) and an ATM machine or a bank branch and also stand in Compatibility (β=.183, P< 0.0001) made a unique and long queues in order to make payment or a withdrawal, statistically significant contribution to the prediction of Mobile payment would help to save time and cost. the Behavioral Intention to Use Mobile payment. With Sig = .287(P > 0.1), Cost does not make a significant Perceived Ease of Use was also found to be the most unique contribution to the prediction of Perceived significant construct affecting Perceived Usefulness Usefulness. Hence, H2a, H3a and H6a are supported (β= .278). Compatibility emerged as the second while H7a was not supported. predictor of Perceived Usefulness and Perceived Ease of Use. From the study, it can be concluded that those To further determine the factors that affect Perceived who have used and are comfortable with Internet Ease of Use of Mobile payment, the impact of banking and Mobile banking and other similar Compatibility and Cost were determined. H7b and technologies (the technology savvy individuals), will H3b confirmed that Compatibility (β = .247, P< likely be the first to try the service and encourage their 0.0005) and Cost (β= .346, P< 0.0005) positively friends, family friends. Trust and Security had a affectPerceived Ease of Use of Mobile significant effect on Perceived Usefulness (β=.163 ), payment.Finally, the result of the single linear issues such as Confidentiality, Integrity, regression between Perceived Ease of Use and Trust Authentication, shows that Trust (β = .306, P< 0.0005) made a unique and statistically significant contribution to the Authorization and Non repudiation affects how the prediction of the Perceived Ease of Use of Mobile consumers perceive the usefulness of the solution. If payment.Hence, H6b is supported. the service providers can provide a service that would not lead to an invasion of privacy, with reduced financial transaction error rates etc. the journey into the cashless Nigeria would be successful.

201

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

REFERENCES

[1] Bagozzi, R.P. & Yi, T. (1988), on the [13] Liu, I., Chen, M.L., Sun, Y.S., Wilbe, D., evaluation of structural equation models, &Kuo C. (2010), Extending the TAM model Journal of the Academy of Marketing to explore the factors that affect Intention to Science, 16(1): 74-94. Use an Online Learning Community, Journal [2] Carter, L., &Be´langer, F. (2005). The of Computer and Education, 54: 600-610. Utilization of e-government services: Citizen [14] Mahdi, S. and Mehrdad, A. (2010), E- trust, innovation and acceptance factors. Banking in Emerging Economy: Empirical Information Systems Journal, 15(1): 5-25. Evidence ofIran, International Journal of [3] CBN (2011), Regulatory Framework for Economics and Finance, 2(1): 201-209 Mobile Payment Services in Nigeria. [15] Mallat, N. (2007). Exploring Consumer [4] Chau, P.Y.K., Hu, P.J.H. (2001). Information Adoption of Mobile Payment: A qualitative technology acceptance by individual study,Journal of strategic Information professionals: a model comparison approach, Systems. Decision. Science. 32 (4): 699–719. [16] Pikkarainen, T., Pikkarainen, K., Karjaluoto, [5] Cheah, C.C., Chuan, A., Sim, J.J. Oon, K.H. H., Pahnila, S. (2004). Consumer acceptance & Tan, B.I. (2011), Factors Affecting of online banking: An extension of the Malaysian Mobile Banking Adoption: An technology acceptance model, Internet Empirical Analysis, International Journal of Research, 14(3): 224-235. Network and Mobile Technologies. ISSN [17] Robinson L. (2009), A Summary of Diffusion 2229-9114 2(3). of [6] Dahlberg, T. (2008). Past, present and future Innovations.http://www.enablingchange.com. of mobile payments research: A literature au/Summary_Diffusion_Theory.pdf (Access review. Electronic Commerce Research and online, May 2012) Applications. 7(2): 165-181. [18] Rogers E.M, (1983).Diffusion of innovations. [7] Davis (1985). A Technology Acceptance Free Press: New York. Model for empirically testing new end-user [19] Rokhman Ali. (2011). E-Government Information Systems: theory and results Adoption in Developing Countries; the Case unpublished dissertations. of Indonesia.Journal of Emerging Trends in [8] Davis, F.D., Bagozzi, R.P., Warshaw, P.R. Computing and Information Sciences, 2(5), (1989). User acceptance of computer 228-233. technology: a comparison of two theoretical [20] Shneiderman, B. (2000), Designing trust into models, Management Science. 35 (8): 982– online experiences, Communications of ACM, 1003. 43:34-40. [9] Dube, T., Chitura, T., Chitura, T. and [21] Taylor, S & Todd, P. A. (1995). Langton, R. (2009),Adoption and Use of Understanding information technology usage: Internet Banking in Zimbabwe: An A test of competing models. Information Exploratory Study, Journal of Internet Systems Research, 6(2): 144–176. Banking and Commerce, 14(1). [22] Venkatesh, V., Davis, F. D. (2000). A [10] Green I.F.R. (2005), The Emancipatory theoretical extension of the technology potential of a new Information System and its acceptance model: Four longitudinal field effect on Technology Acceptance, studies, Management Science, 46(2): 186- Unpublished Ph.D thesis, Pretoria: University 204. of Pretoria.. [23] Wu Jen-Her, Wang Shu-Ching, Lin Li-Min. [11] Lee, Y.-H., Hsieh, Y.-C., & Hsu, C.-N. (2007). Mobile computing acceptance factors (2011). Adding Innovation Diffusion Theory in the healthcare industry: A structural to the Technology Acceptance Model: equation model. International journal of Supporting Employees' Intentions to use E- medical informatics 76,66–77. Learning Systems. Educational Technology & Society, 14 (4): 124–137. [12] Legris, P., Ingham, J. &Colerette, P. (2003). Why do people use information technology? A critical review of the technology

acceptance model. Information and Management, 40, 191-204.

202

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Internet Banking Authentication Methods in Nigeria Commercial Banks

1O.B. Lawal Computer Science Department, Olabisi Onabanjo University Consult, Ibadan, Nigeria [email protected]

A. Ibitola Department of Computer and Information Science Lead City University Ibadan, Nigeria

O.B. Longe Department of Computer Science University of Ibadan Ibadan, Nigeria [email protected]

1Corresponding Author: [email protected]

ABSTRACT The Electronic banking and payments services of commercial banks are recognised by the Central Bank of Nigeria (CBN). Despite the early stage of electronic banking in Nigeria, banks are already offering various financial services through the internet. In order to protect customers’ vital information and identities over the internet, necessary and standard multifactor authentication measures should be in place to avoid financial losses. The purpose of this study is to find out the multifactor authentication (MFA) methods used by the banks, evaluate the type of security mechanism adopted and develop security measures to reliably authenticate customers remotely accessing their Internet-based financial services. The study addressed conducting risk-based assessments and customer awareness program. The study was conducted on all the twenty (20) currently operating commercial banks in Nigeria.

Keywords: Two-factor authentication, internet banking, authentication factor, strong authentication, web security . African Journal of Computing & ICT Reference Format O.B. Lawal, A. Ibitola & O.B. Longe (2013). Internet Banking Authentication Methods in Nigeria Commercial Banks. Afr J. of Comp & ICTs. Vol 6, No. 1. Pp 203-

1. INTRODUCTION

An authentication factor is a piece of information and Financial institutions engaging in any form of Internet process used to authenticate or verify the identity of a banking should have effective and reliable methods to person or other entity requesting access under security authenticate customers. An effective authentication system constraints[1]. Multifactor authentication (MFA) is a is necessary for compliance with requirements to protect system where in two or more different factors are used customer information, to prevent money laundering and in conjunction to authenticate[1]. Using more than one terrorist financing, to reduce fraud, to inhibit identity theft, factor is sometimes called “strong authentication”. The and to promote the legal enforceability of their electronic process that solicits multiple answers to challenge agreements and transactions. The risks of doing business questions as well as retrieves ‘something you have’ or with unauthorized or incorrectly identified persons in an ‘something you are’ is considered multifactor[2].True Internet banking environment can result in financial loss multifactor authentication requires the use of solution and reputation damage through fraud, disclosure of from two or more of the three categories of factors. customer information, corruption of data, or unenforceable Using multiple solutions from the same category would agreements. not constitute multifactor authentication [2].

203

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Wikipedia, the online free encyclopaedia defines internet 2.1 Standards on Protocols banking as “Online banking (or Internet banking or E- The CBN guidance[1] states that “Banks must take banking) allows customers of a financial institution to additional steps to ensure that whilst the web ensures conduct financial transactions on a secure website operated global access to data enabling real time connectivity to by the institution, which can be a retail or virtual bank, the bank’s back-end systems, adequate measures must credit union or building society”. Also, Hazell and be in place to identify and authenticate authorized users Raphael[8] described it as “a number of ways in which while limiting access to data as defined by the Access customers can access their banks without having to be Control List”. Banks are required to ensure that physically present at a bank branch. unnecessary services and ports are disabled. In line with the CBN[1] guidance on “Standards on Authentication schemes according to Sumathi and Protocol”, banks adopt the current and reliable security Esakkirajan[6] are the mechanisms that determine whether measures to authenticate and protect customers’ a user is who he or she claims to be. Authentication can be information while transacting on the websites. carried out at the operating system level or by the RDBMS. The database administrator creates for every user an Oceanic bank[13], on its website, says “Security for individual account or user name[7]. In addition to these communications and transactions over the internet is accounts, users are also assigned passwords. important for both Oceanic Bank and our customers, and we'd like to let you know that the Internet Banking 2. COMPLIANCE REGULATION security system has been selected by us following extensive research”. While the Internet is generally an On August 2003, the Central Bank of Nigeria (CBN) [1] unsecure network, it may be made secure through the issued guidance entitled “Guidelines on Electronic Banking implementation of process controls and infrastructure in Nigeria”. The Guidance focused on future conduct of components[13]. financial institutions (the commercial banks) in e-banking and electronic payments delivery. This guidance applies to 2.2 Standards on Delivery Channels both retail and commercial customers and does not endorse In line with the CBN requirement, standards are placed any particular technology[1]. Financial institutions (banks) on the banks’ delivery channels. Here are the major should use this guidance when evaluating and ones: implementing authentication systems and practices whether they are provided internally or by a service provider. i. Mobile Telephony Although this guidance is focused on the risks and risk Mobile phones are increasingly being used for financial management techniques associated with the Internet services in Nigeria. Banks are enabling the customers to delivery channel, the principles are applicable to all forms conduct some banking services such as account inquiry of electronic banking activities. and funds transfer[1]. Therefore the following guidelines apply: CBN recommend a minimum of two-factor authentication a. Networks used for transmission of financial data process for all user access to the services provided which must be demonstrated to meet the requirements could be high-risk transactions involving access to specified for data confidentiality, integrity and non- customer information or the movement of funds to other repudiation. parties [1]. Financial institutions offering Internet-based products and services to their customers should use b. An audit trail of individual transactions must be kept. effective methods to authenticate the identity of customers using those products and services. The authentication ii. Automated Teller Machines (ATM) techniques employed by the financial institution should be In addition to guidelines on e-banking in general, the appropriate to the risks associated with those products and following, but few of the, specific guidelines apply to services. Account fraud and identity theft are frequently the ATMs: result of single-factor (e.g., ID/password) authentication exploitation. Where risk assessments indicate that the use a. Networks used for transmission of ATM transactions of single-factor authentication is inadequate, financial must be demonstrated to meet the guidelines specified institutions should implement multifactor authentication, for data confidentiality and integrity. layered security, or other controls reasonably calculated to mitigate those risks[1]. b. In view of the demonstrated weaknesses in the magnetic stripe technology, banks should adopt the chip (smart card) technology as the standard. For banks that have not deployed ATMs, the expectation is that chip based ATMs would be deployed.

204

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

iii. Internet Banking Another protocol for transmitting data securely over the Banks should put in place procedures for maintaining the web that the banks employed is Secure HTTP (S- bank’s Web site which should ensure the following[1]: HTTP). It is a modified version of the standard HTTP protocol. By convention, web page that requires an SSL - Banks must ensure that the Internet Service Provider (ISP) connection starts with https, instead of http[15]. has implemented a firewall to protect the bank’s Web site where outsourced. The study discovered that two modes of encryption are - Banks should ensure that installed firewalls are properly in use among Nigerian commercial banks, they are configured and institute procedures for continued 128bit and 256bit SSL. While some banks, such as monitoring and maintenance arrangements are in place. Zenith, GTB, Oceanic and some others uses the 128-bit - Banks should ensure that summary-level reports showing SSL, few others such as Skye bank and Standard web-site usage, transaction volume, system problem logs, Chartered uses 256bit SSL. This can be recognised at and transaction exception reports are made available to the the address bar which starts with 'https'. Also, a bank by the Web administrator. padlock symbol () will be noticed at the bottom of - Web site information and links to other Web sites should the browser[13]. This encryption technology ensures be verified for accuracy and functionality. that data passing between customer computer and the bank is secure and that customer accounts cannot be 3. INTERNET BANKING SERVICE accessed by anyone else online[16].

Apart from the conventional banking practice where ii. Digital Certificate customers are to be physically present at the bank branch, Connolly and Berg[7] defines digital certificate as an information technology has made a new way of banking attachment to an electronic message used for security known as online banking (or internet banking) [1]. This is a purposes, most commonly to verify that a user sending service whereby customers carry on banking transactions a message is who he or she claims to be, and to provide from the comfort of their homes or offices on the internet the receiver with the means to encode a reply. For using personal computer (PC). Interested customers are compliance and security reasons all the banks applied profiled and given a set of log-in detail and signs for digital certificate to send encrypted (username/password), as first factor authentication. With messages. this authentication, customers would be able to do account enquiry and view transaction history. In the attempt to Digital certificate authentication is generally considered make a third party transfer and or third party payment, on one of the stronger authentication technologies, and the internet, another password would be request from the mutual authentication provides a defence against user to verify the genuineness and safety of the customer. phishing and similar attacks[7]. The second password request is the second factor authentication which this study is based on[12]. The use of shared secrets such as digital images is another technique. An image recognition and selection 3.1 SECURITY MECHANISMS is used to identify the genuineness of the customer. This There are general security mechanisms for database method is in use at Enterprise bank website[18]. systems. However, the increasing accessibility of databases in the public internet and private intranets requires a re- iii. Firewall analysis and extension of the approaches, Connolly and When the Web server has to be connected to an internal Berg[7]. There are various identified mechanisms that are network, for example to access the company database, employed by many organisations such as the banks, but for firewall technology can help to prevent unauthorised the purpose of this study, just encryption, digital certificate access, provided it has been installed and maintained and firewalls are evaluated. Other security mechanisms in a correctly[18]. A firewall is a system designed to web environment are proxy server, Kerberos, secure prevent unauthorised access to or from a private electronic transactions (SET), Java Security, Active X network. Following this, it was gathered from the study security etc. that all the banks install robust firewalls to protect their internal systems (intranet) and customer’s information i. Encryption against intrusion from the internet. [19] Secure Socket Layer (SSL) encryption is a secure communication protocol that encrypts client information 4. ADOPTION during transmission over the Internet. It is one of the Nigeria’s slow adoption of electronic banking practice strongest encryption technologies available today, is rapidly changing for the better[21] According to providing server authentication, and ensuring that all data Adeyemi [22] in Aderonke and Charles[21], the transferred over the Internet is encrypted to protect against awareness of electronic payments in Nigeria is it being disclosed to eavesdroppers. It also ensures that any increasing and it accounted for N360 billion worth of attempt by hackers to tamper with the information will be transaction in 2008[22]. detected[14].

205

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

In securing customer data through access control, it is The security of shared secret processes can be enhanced assumed that all bank customers fall in the same user with the requirement for periodic change. Shared group. They all can perform similar operations after gaining secrets that never change are described as “static” and access to the bank’s domain through the internet. the risk of compromise increases over time. The use of Customers on internet banking platform can do balance multiple shared secrets also provides increased security enquiry, check transaction details, make payments and because more than one secret must be known to transfer fund within the same bank. Now customers can authenticate[2]. transfer fund from their bank account to any other bank in Nigeria through the Nigeria Inter-Bank Settlement ii. Tokens System’s (NIBSS) Nigeria Electronic Fund Transfer (NEFT).

5. AUTHENTICATION METHODS

There are a variety of technologies and methods financial Fig. 1: Tokens institutions can use to authenticate customers[4]. These methods include: Use of customer passwords, Personal Tokens are physical devices (something the person has) identification numbers (PINs), Digital certificates using a and may be part of a multifactor authentication scheme. public key infrastructure (PKI), Physical devices such as The hardware consists of a key-fob with an LCD screen smart cards, One-time passwords (OTPs), Use of “tokens” on it. A code is displayed on the screen and changes such as USB plug-ins, Transaction profile scripts, frequently, usually every 60 seconds. The device is Biometric identification and others[4]. The authentication generating keys based on a 128-bit encryption seed. methods adopted by Nigerian banks are passwords, PINs, When this number is fed to a server that has a copy of tokens and One-Time passwords. that seed, it is used as an additional verification to the

other login data[2]. The level of risk protection afforded by each of these techniques varies. The selection and use of authentication There are three general types of token: the USB token technologies and methods should depend upon the results device, the smart card, and the password-generating of the financial institution’s risk assessment process[2]. token. It was gathered from the study that only the

password generating token is in used by the banks[9]. Existing authentication methodologies involve three basic

“factors”: Password-Generating Token • Something the user knows (e.g., password, PIN); A password-generating token produces a unique pass- • Something the user has (e.g., ATM card, smart card, code, also known as a one-time password each time it is token); and used. The token ensures that the same OTP is not used • Something the user is (e.g., biometric characteristic, such consecutively. The OTP is displayed on a small screen as a fingerprint). on the token[2]. The customer first enters his or her

user name and regular password (first factor), followed Authentication methods that depend on more than one by the OTP generated by the token (second factor). The factor are more difficult to compromise than single-factor customer is authenticated if (1) the regular password methods[5]. Accordingly, properly designed and matches and (2) the OTP generated by the token implemented multifactor authentication methods are more matches the password on the authentication server. A reliable and stronger fraud deterrents. For example, the use new OTP is typically generated every 60 seconds—in of a logon ID/password is single-factor authentication (i.e., some systems, every 30 seconds. This very brief period something the user knows); whereas, an ATM transaction is the life span of that password. OTP tokens generally requires multifactor authentication: something the user last 4 to 5 years before they need to be replaced. possesses (i.e., the card) combined with something the user knows (i.e., PIN). A multifactor authentication Password-generating tokens are secure because of the methodology may also include “out–of–band” controls for time-sensitive, synchronized nature of the risk mitigation[5]. authentication. The randomness, unpredictability, and

uniqueness of the OTPs substantially increase the i. Shared Secrets difficulty of a cyber thief capturing and using OTPs Shared secrets (something a person knows) are information gained from keyboard logging[2]. The two elements that are known or shared by both the customer and aforementioned methods of 2-factor authentications the authenticating entity. Passwords and PINs are the best (2FA) are ones basically in use by commercial banks in known shared secret techniques but some new and different Nigeria. There are others methods in use globally such types are now being used as well[2]. as: USB Token devices, Smart Cards, Biometrics, Out-

of-Band Authentication and Mutual Authentication[2].

206

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

Table 1: First Factor Authentication and Security Mechanisms Employed by Nigerian commercial banks S/N Log-in Pass BANKS word Security Encryption Online Certificate Fire wall (128 bit) 1 Access Bank     2 Citibank     3 Diamond Bank     4 Ecobank     5 Enterprise Bank     6 FCMB     7 Fidelity Bank     8 First Bank     9 GTBank     10 Keystone Bank     11 Main Street Bank     12 Skye Bank  *   13 Stanbic IBTC     14 Standard Chartered  *   15 Sterling Bank     16 Union Bank     17 UBA     18 Unity Bank     19 Wema Bank     20 Zenith Bank     * 256bit SSL

Table 1 above shows list of banks, the first-factor authentication (which is the customer username and password), and security mechanisms employed to safeguard banks’ resources and customers’ identities. A combination of encryption modes, digital certificates and robust firewalls are employed by all the banks in compliance. Also, from the above table, all banks adopted a 128bit SSL encryption except for Skye and Standard Chartered banks with higher 256bits SSL encryption security modes.

Table 2: Second Factor Authentication Methods Adopted by the commercial banks S/N BANKS Hardware Token PIN 1 Access Bank  2 Citibank  3 Diamond Bank  4 Ecobank  5 Enterprise Bank  6 FCMB  7 Fidelity Bank  8 First Bank  9 GTB  10 Keystone Bank  11 Main Street Bank  12 Skye Bank  13 Stanbic IBTC  14 Standard Chartered  15 Sterling Bank  16 Union Bank  17 UBA  18 Unity Bank  19 Wema Bank  20 Zenith Bank  TOTAL 10 10 Table 2 above shows list of banks and the 2FA methods adopted. Banks using the hardware token are 10 (50%) while PIN using banks are also 10 representing 50%. 207

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

6. RISK ASSESSMENT However, financial institutions should assess the adequacy of such authentication techniques in light of new or The implementation of appropriate authentication changing risks such as phishing, pharming, malware, and methodologies should start with an assessment of the the evolving sophistication of compromise techniques[3]. risk posed by the institution’s Internet banking Where risk assessments indicate that the use of single- systems. The risk should be evaluated in light of the factor authentication is inadequate, financial institutions type of customer (e.g., retail or commercial); the should implement multifactor authentication, layered customer transactional capabilities (e.g., bill payment, security, or other controls reasonably calculated to wire transfer, loan origination); the sensitivity of mitigate those risks. customer information being communicated to both the The risk assessment process should: institution and the customer; the ease of using the  Identify all transactions and levels of access communication method; and the volume of associated with Internet-based customer transactions[4]. products and services;  Identify and assess the risk mitigation An effective authentication program should be techniques, including authentication implemented to ensure that controls and authentication methodologies, employed for each transaction tools are appropriate for all of the financial type and level of access; and institution’s Internet-based products and services.  Include the ability to gauge the effectiveness of Authentication processes should be designed to risk mitigation techniques for current and maximize interoperability and should be consistent changing risk factors for each transaction type with the financial institution’s overall strategy for and level of access. Internet banking and electronic commerce customer services. 7. OPENING ACCOUNT AND CUSTOMER The level of authentication used by a financial VERIFICATION institution in a particular application should be appropriate to the level of risk in that application[4]. A With the growth in electronic banking and commerce, comprehensive approach to authentication requires financial institutions should use reliable methods of development of, and adherence to, the institution’s originating new customer accounts online. Moreover, information security standards, integration of customer identity verification during account opening authentication processes within the overall information is required and is important in reducing the risk of security framework, risk assessments within lines of identity theft, fraudulent account applications, and businesses supporting selection of authentication tools, unenforceable account agreements or transactions. and central authority for oversight and risk Potentially significant risks arise when a financial monitoring[2]. This authentication process should be institution accepts new customers through the Internet consistent with and support the financial institution’s or other electronic channels because of the absence of overall security and risk management programs. the physical cues that financial institutions traditionally use to identify persons[1]. The method of authentication used in a specific Internet application should be appropriate and reasonable, from a One method to verify a customer’s identity is a business perspective, in light of the reasonably foreseeable physical presentation of a proof of identity credential risks in that application[2]. Because the standards for such as a driver's license international passport or implementing a commercially reasonable system may national ID card[1]. Similarly, to establish the validity change over time as technology and other procedures of a business and the authority of persons to perform develop, financial institutions and technology service transactions on its behalf, financial institutions providers should develop an ongoing process to review typically review articles of incorporation, business authentication technology and ensure appropriate changes credit reports, board resolution identifying officers and are implemented[4]. authorized signatories, and other business credentials. However, in an Internet banking environment, reliance The study agrees with the CBN which consider single- on these traditional forms of paper-based verification factor authentication, as the only control mechanism, to be decreases substantially[1]. Accordingly, financial inadequate for high-risk transactions involving access to institutions need to use reliable alternative methods. customer information or the movement of funds to other parties[3]. Single-factor authentication tools, including passwords and PINs, have been widely used for a variety of Internet banking and electronic commerce activities, including account inquiry, bill payment, and account aggregation.

208

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

8. MONITORING AND REPORTING Methods to evaluate a program’s effectiveness include tracking the number of customers who report fraudulent Monitoring systems can determine if unauthorized attempts to obtain their authentication credentials (e.g., access to computer systems and customer accounts has ID/password), the number of clicks on information occurred[2]. A sound authentication system should security links on Web sites, the number of statement include audit features that can assist in the detection of stuffers or other direct mail communications, the amount fraud, money laundering, compromised passwords, or of losses relating to identity theft, etc[2]. other unauthorized activities. The activation and maintenance of audit logs can help institutions to The study found out that all banks are making efforts to identify unauthorized activities, detect intrusions, educate their customers on how to handle any suspicious reconstruct events, and promote employee and user attempt on their financial details; to ignore any mail accountability. In addition, financial institutions requesting for their PIN and or card details as the bank should report suspicious activities to appropriate would not for any reason request for them, to not enter the regulatory and law enforcement agencies as required bank’s website from links from their email boxes, to by the Bank Secrecy Act[3]. access the internet banking portal from a designated web address. According to the CBN guidance[1], under the section Reporting Requirements states that: 10. RECOMMENDATION a. Banks are required to render separate returns on (a) According to the CBN guidance[1], banks should their e-banking activities to appropriate regulatory introduce logical access controls over ICT authorities as prescribed by the CBN from time to infrastructure deployed. Controls instituted by banks time. should be tested through periodic Penetration Testing, which should include but should not be b. Cases of frauds and forgeries relating to e-banking limited to; should be highlighted in the returns on frauds and a. Password guessing and cracking forgeries. b. Search for back door traps in programs. c. Attempts to overload the system using Banks should rely on multiple layers of control to prevent Ddos (Distributed Denial of Service & DoS fraud and safeguard customer information[1]. Much of (Denial of Service) attacks. this control is not based directly upon authentication. For d. Check if commonly known vulnerabilities example, a financial institution can analyze the activities in the software still exist. of its customers to identify suspicious patterns. Financial e. Banks may for the purpose of such institutions also can rely on other control methods, such as Penetration Testing employ external establishing transaction limits that require manual experts. intervention to exceed a preset limit of amount[1]. f. Continuous and regular customer awareness program to educate Adequate reporting mechanisms are needed to promptly customers. inform security administrators when users are no longer authorized to access a particular system and to permit the A further study to evaluate the reliability and timely removal or suspension of user account access[2]. effectiveness of each of the two most used 2-factor Furthermore, if critical systems or processes are authentication methods, that is, the hardware token outsourced to third parties, management should ensure and the PIN. that the appropriate logging and monitoring procedures are in place and that suspected unauthorized activities are 11. CONCLUSION communicated to the institution in a timely manner. An independent party (e.g., internal or external auditor) Financial institutions offering Internet-based products should review activity reports documenting the security and services should have reliable and secure methods administrators’ actions to provide the necessary checks to authenticate their customers. The level of and balances for managing system security. authentication used by the financial institution should be appropriate to the risks associated with those 9. CUSTOMER AWARENESS products and services. Financial institutions should conduct a risk assessment to identify the types and Banks have made, and should continue to make, efforts to levels of risk associated with their Internet banking educate their customers. Because customer awareness is a applications. Where risk assessments indicate that the key defence against fraud and identity theft, financial use of single-factor authentication is inadequate, institutions should evaluate their consumer education financial institutions should implement multifactor efforts to determine if additional steps are necessary[1]. authentication, layered security, or other controls Management should implement a customer awareness reasonably calculated to mitigate those risks. program and periodically evaluate its effectiveness.

209

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

The Central Bank of Nigeria (CBN) consider single- factor authentication, as the only security control [10] ttp://httpd.apache.org/docs/2.2/howto/auth.html mechanism, to be inadequate in the case of high-risk [11] http://www.naijatechguide.com/2010/02/internet transactions involving access to customer information -banking-security-tips.html or the movement of funds to other parties. It was [12] http://www.cenbank.org/Supervision/Inst- discovered from the study that all Nigerian banks have DM.asp adopted and implemented the 2FA methods as [13] https://ibank.oceanicbank.com/corp/web/L001/ mandated by the CBN and to meet international corporate/jsp/user/onlinesecurity.htm standards. [14] http://www.zenithbank.com/ibanksecurity.cfm [15] https://ebanking.sterlingbankng.com/IBS/index.j The success of a particular authentication method sp depends on more than the technology. It also depends [16] https://ibank.fidelitybankplc.com on appropriate policies, procedures, and controls. An [17] http://www.gtbank.com/personalbanking/ways- effective authentication method should have customer to- bank/internetbanking acceptance, reliable performance, scalability to [18] http://web.entbanking.com accommodate growth, and interoperability with [19] https://ibank.oceanicbank.com/corp/web/L001/ existing systems and future plans. corporate/jsp/user/onlinesecurity.htm [20] http://hackaday.com/2009/10/20/two-factor- Three methods were used to gather information for the authentication-using-a-hardware-token/ survey; (1) information from the banks’ websites, (2) [21] Adesina and Charles. An Emperical telephones calls to the e-banking units of some of the Investigation of the levels of User’ Acceptance banks, and (3) enquiring from the customer service of E-Banking in Nigeria. Journal of Internet officers in some bank’s branches where information banking and Commerce. Vol. 15, No. 1. April could not be gotten from any of the previous two 2010. methods. [22] Ayo, C. K. Adebiyi A. A., Fatudimu I.T., Ekong O.U. (2008). Framework for e-Commerce REFERENCES Implementation: Nigeria a Case Study, Journal of Internet Banking and Commerce, August [1] Central Bank of Nigeria: Guidelines on 2008, vol. 13, no.2. Electronic Banking in Nigeria, August, 2003 [2] Federal Financial Institutions Examination Council (FFIEC) agencies, Authentication in an Electronic Banking Environment, 2001. [3] FFIEC Information Technology Examination Handbook, Information Security Booklet, December 2002 [4] FFIEC Information Technology Examination Handbook, E-Banking Booklet, August 2003 [5] Basel Committee on Banking Supervision, “Risk Management Principles for Electronic Banking”. July 2003. Bank for International Settlements. [6] Sumathi S. and Esakkirajan S. Fundamentals of Relational Database Management Systems, Springer, 2007. [7] Database Systems. A practical Approach to Design, Implementation and Management. Addison Wesley. Fourth Edition, 2005. [8] Paul Hazell and Ziad Raphael, Internet Banking: Disruptive Or Sustaining Technology? Field Project Report submitted to Harvard Business School Boston, MA. 2001. [9] Adeoye O. S. (2012) Evaluating The Performance Of Two-Factor Authentication Solution In The Banking Sector. Ijcsi International Journal Of Computer Science Issues, Vol. 9, Issue 4, No 2, July 2012. Issn (Online): 1694-0814.

210

Vol 6. No. 1 March, 2013 African Journal of Computing & ICT

211