Predicting Student’s Academic Performance in a Learning Management System Platform Using Ensemble Techniques

Learning Management Systems (LMS) facilitate systematic teaching in various learning environments, ensuring continuity regardless of disruptions. These platforms allow students to attend classes, participate in discussions, and submit weekly tasks, mitigating concerns about class disturbances. However, research indicates that while LMS benefits many students, others face significant challenges due to various factors. This study explores how conventional factors—such as personal information, personality type, study habits, digital resources, and internet accessibility—impact academic performance among students in the College of Computing Studies and the College of Social Work and Community Development. Furthermore, this research aims to develop a new ensemble-based model that leverages bagging and boosting techniques to enhance predictive accuracy. The study adopts an experimental research design to identify the most effective model. Findings indicate that the boosting model outperforms other ensemble methods in predicting academic success. Additionally, hypothesis testing using ANOVA confirms a significant positive relationship between these factors and academic performance. However, this study recommends future research with a larger and more balanced dataset to improve the reliability of findings. Future work should explore advanced data balancing techniques, such as Synthetic Minority Over-sampling Technique (SMOTE), to address dataset imbalances and enhance model generalization. Additionally, optimizing ensemble models through hyperparameter tuning using Grid Search, Random Search, or Bayesian Optimization can refine performance and predictive accuracy.