Predicting Correctness of Problem Solving in ITS with a Temporal Collaborative Filtering Approach

Collaborative filtering (CF) is a technique that utilizes how users are associated with items in a target application and predicts the utility of items for a particular user. Temporal collaborative filtering (temporal CF) is a timesensitive CF approach that considers the change in user-item interactions over time. Despite its capability to deal with dynamic educational applications with rapidly changing user-item interactions, there is no prior research of temporal CF on educational tasks. This paper proposes a temporal CF approach to automatically predict the correctness of students’ problem solving in an intelligent math tutoring system. Unlike traditional user-item interactions, a student may work on the same problem multiple times, and there are usually multiple interactions for a student-problem pair. The proposed temporal CF approach effectively utilizes information coming from multiple interactions and is compared to i) a traditional CF approach, ii) a temporal CF approach that uses a sliding-time-window but ignores old data and multiple interactions and iii) a combined temporal CF approach that uses a sliding-time-window together with multiple interactions. An extensive set of experiment results show that using multiple-interactions significantly improves the prediction accuracy while using sliding-time-windows doesn’t make a significant difference.