A new method for learnersʹ knowledge level estimation with personalized initial probability of knowledge

Modeling learners based on the Bayesian Network (BN) can help us in coming to a conclusion regarding the learner’s state of knowledge. Determining the initial probability for each root node in BN and also conditional probability is necessary to model learners in the Bayesian Network. Determining the personalized initial probability of knowledge is one of the important cases in modeling learnersʹ knowledge to improve adaptivity and intelligence of E-learning systems. In this paper, in order to solve the problem of personalized initial knowledge, the Bayesian Knowledge Tracing Algorithm (BKNT) is employed. Nodes in the Bayesian Network are the concepts of knowledge domain Ontology and the relation between concepts are prerequisite relations. Initial knowledge can be obtained from performing BKNT on a test before learning (Pretest) and conditional probabilities of BN can be learned from a dataset that have been previously obtained from a group of learners. Finally, by using the initial probabilities and conditional probabilities, the estimation of learnersʹ knowledge level on the next concepts of Ontology can be performed. To assess the used method, learners answered a test including a question per each next concept and selected choices with dedicated standard percentage to each choice.Results of estimation are acceptable and they are an improvement in adaptive E-learning systems