No abstract

This paper uses Expectation Maximization (EM) to learn the hidden characteristic of a student's mastery of mathematical skills. In particular, we build a Bayesian network (BN) based on student pretests of problems using 12 different skills and then run inference to predict a student's individual mastery of each skill. We utilize the Bayesian Information Criterion (BIC) to evaluate different skill models. This learned knowledge of a student's initial skill levels is essential to the overall effectiveness of the Intelligent Tutoring System (ITS).