Abstract—This paper presents an agent based Intelligent Tutoring System (ITS). The agent plays the role of an advisor for an e-learner to facilitate the learner to achieve his learning objective. It provides advices to assist an e-learner while solving problems that are normally provided by human experts. A Bayesian network is employed to construct the possible solution states of a problem with the statistics of mistakes; those can be committed by a learner in their problem solving process. The agent will collect statistical data from Bayesian network on learner’s mistakes and the way in which one learner may commit a mistake. On the basis of the statistical data on the learner’s behaviour, especially with respect to his tendency to commit an error, the agent anticipates the point of difficulty during a problem solving session and accordingly guides the learners to solve the given problem free of errors. The Bayesian network is trained with training data, prior information (e.g., expert knowledge, casual relationships, and estimated graph topology or network structure) and the parameters of the joint probability distribution. Index Terms—Intelligent advisory system (ias), intelligent agent, Bayesian network. I. INTRODUCTION An enormous development in the field of information technology (IT) has opened up some e-systems parallel with traditional ongoing systems such as e-banking systems parallel with banking system, e-business, e-education, e-learning etc. This development has also opened up immense possibilities in the contemporary education and training arena. The traditional classroom based teaching learning systems are augmented by various kinds of e-learning systems e.g. Intelligent Tutoring Systems (ITSs), Learning Management Systems (LMS), and Virtual Laboratory etc. These all software systems can significantly improve their performance if they could adapt to the emotional state of the user, for example if Intelligent Tutoring Systems, ATM’s, Ticketing machines could recognise when users were confused, frustrated, angry or whether the learner in ITSs understood the learning materials or not then they could guide the user back to remedial help systems so improving the service. A tremendous amount of research activities are currently going on in developing these software systems. Many researchers now feel strongly that ITSs would be significantly enhanced if computers could adapt to the emotions of students [1]. This idea has spawned Manuscript received September 15, 2011; revised October 28, 2011 The authors are with the department of Computer Science and Engineering from NITTTR’ Kolkata.india (e-mail: pratyay_kuila@yahoo.com.) the developing field of Affective Tutoring Systems (ATSs): ATSs are ITSs that are able to adapt to the affective state of students [2]. The term ‘‘affective tutoring system’’ can be traced back as far as Rosalind Picard’s book Affective Computing in 1997. In traditional tutoring system student advising is a very important issue. Here, we can estimate the student state, understanding level and according to these we can advice them. However student advising is not given enough attention in Intelligent Tutoring Systems (ITS) [3], [4], [5]. Current researchers in the field of ITSs are at investigation of how to make a computer- based tutors more flexible, autonomous and adaptive to evaluate each student. In basic ITS one student is evaluated by some multiple choice questions. Here the main drawback is, when one student makes some wrong answer, there is no chance of understanding the proper step, where the student has committed the mistake while solving the problem. Usually, solution of a problem consists of a sequence of steps. A learner may commit a mistake at any step during the solution process. So, if we can identify the proper step, where a student has been committed a mistake then we can advice them according to their mistakes. Student modeling is crucial for an intelligent learning environment to advice a student according to their knowledge and understanding ability. The basic underlying perspective is to consider every student as being unique and advice him according their knowledge, understanding and types of mistake while solving a problem. This paradigm can be built by creation of a user or student model [6]. The main objective of student model is to understand the knowledge level of individual student, how they learn and what their problems are while they learn [7]. Assessing the user state of knowledge and profile requires uncertainty reasoning. Artificial Intelligence has addressed this problem in various ways such as fuzzy logic, Bayesian Networks [8] etc. So, using student model ITS can build an advisory system to advice a student students in their learning process by maintaining an accurate model of a student’s understanding level, current knowledge state which allows more intelligent pedagogical decisions and actions to be happened. Using student modeling and the techniques of artificial intelligence we can build an advisory system to advice a student in their learning process. An advisory system provides the advices and assists for solving problems that are normally provided by human experts. They can be classified as a type of expert systems [9], [10]. Advisory system has the power to make recommendations but not to enforce a decision maker and it does not make decisions but rather help guide the decision maker in the decision-making process. Bayesian Network Based Intelligent Advice Generation for Self-Instructional e-Learner P. Kuila, C. Basak, and S. Roy 280 International Journal of e-Education, e-Business, e-Management and e-Learning, Vol. 1, No. 4, October 2011