<![CDATA[Digital learning environments and Learning Management Systems (LMSs) generate large volumes of time-stamped behavioral traces that can be used to examine how students access resources, navigate course structures, communicate, and approach assessments. Traditional learning-style models often depend on static self-report categories and may not reflect how students actually study in digital courses. This study develops a learning analytics framework for modeling student learning profiles from authentic LMS behavioral traces. The study used a quantitative, non-experimental, longitudinal design based on Canvas LMS interaction data from 15,342 undergraduate students enrolled in 150 large-enrollment courses during the 2023–2024 academic year. More than 500 million raw interaction logs were processed into 24 engineered behavioral features representing temporal engagement, resource access, navigation behavior, interaction activity, and assessment timing. After feature normalization, K-Means clustering was applied, and the optimal cluster solution was selected using the elbow method and average silhouette score. Cluster distinctiveness was examined using one-way analysis of variance, and the association between cluster membership and academic performance category was evaluated using a Chi-squared test. The analysis supported a four-cluster solution. Assessment procrastination and navigation sequentially were the strongest differentiating features.]]>
Copyrights © 2026
