<![CDATA[This study examined whether behavioral analytics derived from a Moodle-based learning management system could help identify early signs of learning disengagement in blended higher education. A quantitative predictive correlational design was used with 305 undergraduate students at Universitas Islam Negeri Raden Intan Lampung, Indonesia. Five behavioral indicators, login frequency, assignment submission timeliness, discussion participation, video completion rate, and LMS interaction breadth, were analyzed alongside scores from a 24-item Learning Disengagement Index (LDI). Multiple linear regression and four machine learning classifiers were evaluated using stratified 10-fold cross-validation. The composite behavioral model was significantly associated with LDI scores (R² = .671, p < .001), with login frequency (β = −.42) and assignment submission timeliness (β = −.38) showing the strongest standardized associations. At the Week 4 checkpoint, Random Forest showed the highest classification performance among the tested algorithms (Accuracy = .832; AUC = .912), followed by Gradient Boosting. These findings suggest that early LMS behavioral traces can provide useful decision-support signals for student support in blended courses. The results should be interpreted as context-specific predictive evidence rather than causal evidence, and local validation is recommended before institutional deployment.]]>
Copyrights © 2026
