Tracking At-Risk Student Groups from Teaching and Learning Activities in Engineering Education

Abstract

Tracking student groups, in particular, at-risk student group is a challenging but meaningful work in a large class of an engineering mathematics course, enabling instructors to ascertain how well students are learning and when they need interventions of their studies during the delivery of teaching and learning activities. In the paper, two unsupervised learning algorithms, hierarchical clustering and k-means clustering, are used and compared with the use of LMS data such as the level of achievements in online class activities, assignments, a mini-project and a mid-term test for tracking at-risk student groups at the end of weeks 3, 5, 7, 9 and 11 in a 13-week semester of an academic year. Notwithstanding the higher accuracy of both clustering, the k-means clustering significantly outperforms the hierarchical clustering in terms of the precision, recall and f-measure at the end of week 11. It is found that the k-means clustering can be employed to track at-risk students with the recall of 0.640 and the f-measure of 0.533 for the initial intervention of their studies by the end of week 7.