Using Clustering Techniques to Understand Student Involvement in an Online Environment

Student engagement in online learning environments is critical in improving educational outcomes and instructional strategies. Previous studies on engagement patterns using online log datasets often focus on interaction frequency, neglecting intensity and comprehensive activity coverage. This study addresses these gaps by introducing a novel approach grounded in the Community of Inquiry (CoI) model to calculate engagement parameters. The research objectives include deriving meaningful engagement metrics, clustering students based on these metrics, and evaluating clustering algorithms to identify the most effective method. The methodology involves processing Moodle log data to extract three key engagement parameters: Number of sessions, session duration, and engagement levels encompassing social and cognitive dimensions. These derived parameter values were then compared to the labels set manually by two raters. High agreement (0.9409 correlation) between these two methods validates the algorithm’s efficiency and reliability in measuring student engagement. Next, clustering algorithms, such as Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Gaussian Mixture Model (GMM), K-means, Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH), etc., are applied to group students, with cluster quality assessed using indices like Davies-Bouldin, silhouette coefficient, and Calinski-Harabasz. The findings reveal that Kmeans and Birch algorithms effectively categorize students, with the CoI-derived engagement parameters proving to be the most influential. These insights highlight the critical role of cognitive and social interactions in engagement and demonstrate the superiority of such methods in discovering patterns in student data. This study provides a robust framework for analyzing student engagement, offering actionable insights for educators to enhance online learning experiences.