<![CDATA[Student retention is a crucial issue in higher education, as low retention rates directly increase the risk of dropout and affect institutional reputation and sustainability. This study aims to compare the performance of single machine learning (ML) models, namely Naïve Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Logistic Regression (LR), Random Forest (RF), and Artificial Neural Network (ANN), with an optimized ensemble learning approach using Voting Classification. The optimization was carried out through weighted soft voting, the selection of complementary model combinations, and handling class imbalance with the SMOTE method. The student dataset was processed through preprocessing techniques and validated using 10-fold cross-validation, with model performance evaluated by accuracy, precision, recall, and F1-score metrics. The results indicate that Soft Voting Classification consistently outperforms individual models, with the KNN_ANN_SVM combination achieving the highest accuracy of 87.55%. However, the experiments also show that the application of SMOTE did not improve performance. The best-performing model has the potential to serve as the basis for developing an Early Warning System to support data-driven interventions in reducing student dropout risk.]]>
Copyrights © 2026
