Collaborative filtering for student performance prediction

Student performance prediction in educational data mining aims at predicting the correctness in performing a given task of a student based on historical data. Accurately predicting the performance allows helpful intervention and useful feedback by instructors to improve students’ learning. This paper presents a framework for tackling the student performance prediction problem using collaborative filtering, which collectively considers behaviour of students as well as the nature and difficulty of questions. Experiments on real-world data of an intelligent tutoring system were conducted to evaluate the effectiveness of our framework.