UniNet: Next Term Course Recommendation using Deep Learning

TL;DR

This paper introduces UniNet, a deep learning model using recurrent neural networks to improve next term course recommendations by capturing the temporal sequence of student grades, outperforming traditional methods.

Contribution

The paper presents a novel RNN-based recommender system that models the chronological order of grades, enhancing course recommendation accuracy over existing techniques.

Findings

Achieved 81.10% AUC using grade data alone

Effective across various GPA levels and course difficulties

Demonstrated improved performance over traditional collaborative filtering

Abstract

Course enrollment recommendation is a relevant task that helps university students decide what is the best combination of courses to enroll in the next term. In particular, recommender system techniques like matrix factorization and collaborative filtering have been developed to try to solve this problem. As these techniques fail to represent the time-dependent nature of academic performance datasets we propose a deep learning approach using recurrent neural networks that aims to better represent how chronological order of course grades affects the probability of success. We have shown that it is possible to obtain a performance of 81.10% on AUC metric using only grade information and that it is possible to develop a recommender system with academic student performance prediction. This is shown to be meaningful across different student GPA levels and course difficulties