Clustering Students Using Pre-Midterm Behaviour Data and Predict Their Exam Performance

Student behaviour should correlate to the course performance. This paper explored different types of clustering algorithms using the pre-midterm student behaviour data. We found meaningful and interpretive results when clustering algorithms generate three clusters. The clusters can be briefly summarized as potential top performance (PTP) students, potential poor performance (PPP) students, and mixed performance (MP) students. We found that PTP students usually submit early and gain a high score, PPP students usually submit late and gain a low score, and MP students usually make most submissions. MP students are hard to cluster. However, we found a good connection between other students' behaviour and performance if we leave out MP students