Pakistan Journal of Science (Vol. 68 No. 2 June, 2016) 239 PERSONALIZED E-LEARNING SYSTEMS: A USER MODELING TECHNIQUE S. Rajper and A. W. Shaikh Department of Computer Science, Shah Abdul Latif University, Khairpur Corresponding Author e-mail: samina.rajper@gmail.com ABSTRACT: For providing the personalization feature to e-learning systems, a Human Computer Interaction HCI technique, User Modeling was considered in system design. This study was undertaken to propose a technique to model the students’ learning styles f or personalized e-learning system. Bayesian Network BNwas used to model the students’ learning styles and Kolb’s learning styles theory was used to know the students’ learning styles and preferences. The objective of the study was to propose a user modeling technique for personalized e-learning systems to understand the e- learners’ requirements and needs to enhance their learning. Using BN, Conditional Probability Tables CPTs were determined for all four learning styles provided by Kolb’s theory. The thresho ld values were used to determine the learning styles of students during an online course. We found that BN technique determined successful as from the Divergers 100%., 75% Assimilator, 50% Accommodators and 75% Convergers were identified accurately. Key words: Personalized e-learning systems, User Modeling, Bayesian Network, Learning Styles. (Received 08-02-2016) Accepted 20-06-2016) INTRODUCTION User Modeling is an essential part of intelligent systems design. Personalized e-learning systems are the intelligent systems which can be characterized as Adaptive Hypermedia Education Systems AHES and Intelligent Tutoring Systems ITS. The AHES have an essential characteristic of adaptivity as per the students’ model (Brusilovsky and Millán 2007). User Modeling is a technique to represent the original users of the system. The AHES can adapt the learning contents and other activities for e-learners as per their user models. For user modeling, the data can be gathered either from observing the users’ behavior or from filling the forms directly from the users. This data is used to model and classify the users to facilitate them accordingly(Brusilovsky 1998; Brusilovsky 2007; Brusilovsky and Millán 2007). Many past studies, i.e.,(Brusilovsky and Millán 2007; Schiaffino, Garcia et al. 2008; Özpolat and Akar 2009; Yang, Hwang et al. 2013; Truong 2016) have mentioned the use of learning styles to model the students for personalized e-learning systems.Many user model techniques have been used to model the students, i.e., Profiles in ITS by(Schiaffino, Garcia et al. 2008), Students’ knowledge level by(Boyle and Encarnacion 1998), Learning styles by (Weber and Specht 1997; Graf 2007). Past research studies have used Felder Silverman’s learning style theory (Felder and Silverman 1988) to adopt the learning styles for personalized e- learning systems, because this learning style theory is easy to be incorporated to understand students’ learning styles on learning management system (Graf 2007). However, some research studies, i.e., (Derntl and Graf 2009; Siraj 2012) have reported unsatisfactory results of personalized e-learning systems in which Felder Silverman’s learning style theory was adopted. Therefore, Kolb’s learning styles model (Kolb 1984) is used during present study to understand the students’ learning styles on learning management system. This study is using probabilistic technique Bayesian Network (Jensen 1996) to model the students / e-learners according to the Kolb’s learning style theory(Kolb 1984) MATERIALS AND METHODS Data Collection: When one wants to model the user, it is required to collect huge data about the user. For modeling students of e-learning system using learning styles, it is required to map the students’ behavior on learning management system with his/her learning style. Therefore, a survey of more than 800 students (e- learners) was conducted to understand and map their behavior with their learning styles. The students’ survey data provided us information about e-learner’s learning styles and their activities on learning management system, i.e.,  Login time for attending online lecture  E-learners’ Material reading behavior  Assignments/Quizzes submission behavior.  Preferred contact person during course learning. The next step was to model the students’ behavior as variables using BN. E-learners’ Modeling using Bayesian Network: For e- learners’ modeling BN was used. The basic equation of Bayesian Network BN was used, mentioned as Eq.1.