Student Performance Analysis in Online and Offline Adaptive Learning Environments using Ensemble Machine Learning

Adaptive learning systems have become increasingly important in modern classrooms as they enable personalized learning experiences that address the diverse needs of students. Traditional teaching approaches, particularly in large class settings, often fail to accommodate individual differences, which can reduce student engagement and widen performance gaps. To address these challenges, this study proposes an ensemble machine learning framework for predicting student performance across both online and offline adaptive learning environments. The proposed Ensemble Voting Regressor (EVR) integrates Random Forest Regression, Ridge Regression, Support Vector Regression, and Decision Tree Regression through a weighted voting mechanism to enhance predictive robustness and accuracy. A comprehensive preprocessing pipeline was employed, including data cleaning, categorical feature encoding, normalization, and dataset partitioning into 80\% training and 20\% testing subsets to ensure reliable model evaluation. Experimental results demonstrate strong predictive performance, achieving accuracy rates of 99.94\% for the online dataset and 99.98\% for the offline dataset, with very low mean absolute error values of 0.0080 and 0.0081, respectively. Feature analysis revealed that study time, prior academic failures, and engagement-related attributes significantly influence student performance. Furthermore, the proposed model provides personalized feedback to support targeted academic interventions. Overall, the findings highlight the effectiveness of data-driven machine learning approaches in enhancing adaptive education, promoting equitable learning outcomes, and supporting informed educational decision-making.