Predicting academic performance via Semi-Supervised learning with constructed campus social network

Abstract

Wide attention has been recently paid to academic performance prediction, due to its potentials of early warning and subsequent in-time intervention. However, there are few studies to consider the effect of social influence at predicting academic performance. The major challenge comes from the difficulty of collecting a precise friend list for students. To this end, we first construct students’ social relationship based on their campus behavior, and then predicts academic performance using constructed social network by semi-supervised learning. We evaluate the proposed algorithm on over 5,000 students with more than 14M behavior records. The evaluation results show the potential value of campus social network for predicting academic performance and the effectiveness of the proposed algorithm.