Comparing Machine Learning Algorithms with Ensemble Model Using Random Oversampling for Predicting Student's Dropouts

Studying at the university can be very difficult for some students, and many struggle to keep up with their academics, resorting to academics dissatisfaction and even dropping out. This study examines the impact of various machine learning models in classifying student dropout likelihood, with a focus on mitigating class imbalance using Random Oversampling (ROS). We compare machine learning algorithms such as Logistic Regression (LR), K-Nearest Neighbor (KNN), Naïve Bayes (NB), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Soft Voting. Our results show that the Soft Voting model achieves the best performance, with an F1 score of 0.93, outperforming other models in predicting dropouts. ROS is found to improve model performance, yet it contributes to slight variability, indicating potential overfitting. These findings support educational institutions in designing targeted interventions for at-risk students, helping to prevent dropout through optimized prediction methods.