Multi-Relational Factorization Models for Predicting Student Performance
Nguyen Thai-Nghe, Lucas Drumond, Tomáš Horváth · 2011
Predicting student performance (PSP) is the problem of predicting how well a student will perform on a given task. It has gained more attention from the educational data mining community recently. Previous works show that good results can be achieved by casting the PSP to rating prediction problem in recommender systems, where students, tasks and performance scores are mapped to users, items and ratings respectively. One of the most prominent approaches for rating prediction which also performs well in PSP is matrix factorization (MF). However, the state-of-the-art MF approaches for PSP only make use of one relationship, that is, between students and tasks or students and skills needed to solve the tasks. In fact each student performs several tasks, and the tasks relate to the skill(s) needed to solve them, while students are also required mastering on the skills that they have learned. In this paper we propose to exploit such multiple relationships by using multi-relational MF methods. Experiments on three large datasets show that the proposed approach can improve the prediction results. 1.