<![CDATA[This study analyzes the effectiveness of resampling techniques and ensemble learning in addressing class imbalance problems in student performance prediction using the xAPI-Edu-Data dataset from the Kalboard 360 LMS. The class imbalance ratio of 1:1.66 leads to bias in traditional classification models toward the majority class. The study evaluates six resampling methods, including hybrid SMOTE-ENN, combined with nine individual classifiers and three ensemble models (bagging, voting, and stacking). Evaluation was conducted using accuracy, precision, recall, and F1-score with stratified 5-fold cross-validation and hyperparameter optimization through GridSearchCV. The results indicate that the combination of SMOTE-ENN with voting and stacking achieved the best performance of 98.18% across all evaluation metrics and significantly improved minority-class recall, demonstrating its effectiveness for developing early warning systems to identify at-risk students.]]>
Copyrights © 2026
