Evaluating Students’ Academic Proficiency using Unsupervised Learning approach for Performance Prediction

Abstract

This research reports the use of unsupervised learning approach, specifically nearest centroid and connectivity clustering, with Principal Component Analysis (PCA) to analyze student academic and behavioral datasets. By revealing hidden patterns without labelled outcomes, this research illustrates the potential combinations of clustering and labelling to identify at-risk learners, moderate learners and those with higher levels of success. The results suggest that unsupervised learning serves as a valuable complement to supervised learning within educational applications. By leveraging data mining strategies, these methods enable timely interventions and promote personalized learning pathways.