Application of Time Decay Functions and the Elo System in Student Modeling

One of the key aspects of educational data mining is estimation of student skills. This estimation is complicated by the fact that students skills change during the use of an educational system. In this work we study two flexible approaches to skill estimation: time decay functions and the Elo rating system. Results of experiments in several different settings show that these simple approaches provide good and consistent performance. We argue that since these approaches have several pragmatical advantages (flexibility, speed, ease of application) they should be considered in educational data mining at least as a baseline approach.