Predictive Modeling of Student Performance in Moodle LMS using Learning Analytics

This research paper explores the use of learning analytics to predict student performance in the Moodle Learning Management System (LMS). It examines the accuracy of various predictive models in determining student success or failure and in predicting final grades. Ten models were evaluated using Moodle student records, which included engagement patterns and academic outcomes. The results showed significant variation in model accuracy. The Gradient Boosting Classifier achieved an accuracy of 96.296% in predicting course outcomes, while Random Forest model demonstrated 92.593% accuracy. For predicting final grades, the Random Forest, Decision Tree, and Gradient Boosting Classifiers all achieved an accuracy of $\mathbf{92.593\%}$, with other models also showing good accuracy. These findings offer valuable insights for teachers and students to identify at-risk students and enhance academic outcomes. Further research is recommended to explore the factors and features that contribute to predictive modeling, including gender-based variations in student success, and to refine models for improved performance.