Educational Data Mining of VLE Interaction Logs for Explaining Student Assessment Outcomes: A K Means Clustering Study

Abstract

Virtual Learning Environments (VLEs) generate extensive traces of learner interaction; however, educators and institutions often lack interpretable analytical approaches to determine which online learning activities meaningfully contribute to assessment success. This study applies Educational Data Mining (EDM) techniques to investigate the relationship between students’ online learning behaviors and assessment outcomes using large-scale VLE interaction data. The Open University Learning Analytics Dataset (OULAD), containing detailed student, assessment, and activity records across multiple course presentations, is used as the empirical basis of the analysis. A consolidated learner–activity–assessment dataset is constructed through relational integration and preprocessing of multiple OULAD tables. K-Means clustering is applied to identify distinct engagement–performance profiles without relying on predefined outcome labels. The analysis results in four clusters with heterogeneous engagement characteristics, among which one cluster emerges as dominant, capturing the highest proportion of meaningful learning interactions and demonstrating a strong association with higher assessment performance. Further analysis examines activity-type distributions across clusters to evaluate the relative influence of different online learning activities. The results indicate that engagement with structured learning resources and core content activities exhibits the strongest relationship with academic success, while activities supporting formative practice and interaction contribute positively when combined with content focused engagement. In contrast, excessive interaction with navigational or supplementary activities shows limited association with improved performance. These findings suggest that engagement quality and activity semantics are more informative indicators of learning effectiveness than overall activity volume. The study provides interpretable, cluster-based evidence to support data-driven instructional decision-making and highlights the importance of guiding students toward high impact learning activities in higher education contexts.