The integration of Learning Management Systems (LMS) in educational institutions has revolutionized teaching and learning processes by enabling digital access, activity tracking, and performance monitoring. However, despite the growing availability of LMS data, many institutions struggle to identify students at risk of academic failure in a timely manner. This study explores the use of data mining techniques to analyze LMS log data and detect patterns indicative of at-risk student behaviour. Using a dataset derived from a university LMS platform, the study focuses on key engagement metrics such as login frequency, time spent on the platform, assignment submission punctuality, quiz scores, and participation in discussion forums. Classification algorithms including Decision Trees and Naïve Bayes, as well as unsupervised clustering methods like k-Means, are employed to analyze the data and predict at-risk profiles. The results reveal significant correlations between digital engagement and academic performance. Students with low login frequency and delayed submissions were more likely to fall into the at-risk category. The predictive models achieved an accuracy rate of over 85%, indicating the reliability of data mining techniques in educational analytics. These findings underscore the potential of LMS log analysis as a proactive tool for educators and academic advisors. By identifying at-risk students early in the term, institutions can implement targeted interventions to improve learning outcomes and reduce dropout rates. This research contributes to the growing field of Educational Data Mining (EDM) and highlights the importance of data-driven decision-making in modern academic environments.