Data-Driven Construction of a Student Model Using Bayesian Networks in an Electrical Domain

The student model is a key component of intelligent tutoring systems since enables them to respond to particular needs of students. In the last years, educational systems have widespread in school and industry and they produce data which can be used to know students and to understand and improve the learning process. The student modeling has been improved thanks to educational data mining, which is concerned with discovering novel and potentially useful information from large volumes of data. To build a student model, we have used the data log of a virtual reality training system that has been used for several years to train electricians. We compared the results of this student model with a student model built by an expert. We rely on Bayesian networks to represent the student models. Here we present the student models and the results of an initial evaluation.