Abstract
<jats:p>In the era of distance education, understanding student behaviours through learning analytics has become crucial for improving educational outcomes. This study analyses log data obtained from a Learning Management System (LMS) to explore the relationship between student engagement patterns and academic performance. Student's interaction with various activity contexts were extracted as features and evaluated through machine learning methods. A Random Forest model, supported with Local Interpretable Model-Agnostic Explanations (LIME) interpretability, was applied to classify outcomes. The model achieved an accuracy of 81.67%, revealing that attendance patterns and exam duration are critical factors influencing success. Findings offer practical implications for instructors and supporting the design of targeted interventions, early warning systems, and personalised feedback mechanisms. Ultimately, the research demonstrates the potential of log-based analytics and Explainable Artificial Intelligence to deliver evidence-driven improvements in distance education and student support.</jats:p>