Clustering Students Based on Online Learning Interactions Using Social Network Analysis

Abstract

One of the key areas of interest within learning analytics is identifying student similarities to support collaborative applications such as score prediction, personalized recommendations, and group formation. Clustering is a prominent method for grouping students based on shared learning behaviors, enhancing peer learning, and fostering communities in online courses. This study introduces an intuitive graphical clustering approach using activity logs that track student engagement with different learning resources. These interactions are modeled as multi-dimensional vectors, and a social network of learners is constructed using cosine similarity. Social network analysis (SNA) is then applied to detect learner communities. The dataset for implementation and evaluation includes activity logs and grades from 792 students in an undergraduate study program. Results indicate that learners in the same clusters have similar interaction patterns and grade point averages (GPAs). Statistical measures, such as silhouette index and root mean square standard deviation (RMSSTD), demonstrate the method’s effectiveness and benchmark its performance against K-means clustering. This approach shows significant potential for uncovering and visualizing implicit learner groups.