School Dropout Prediction with Class Balancing and Hyperparameter Configuration
摘要
School dropout and academic underachievement have significant effects on economic growth and employment in society. This phenomenon impacts not only the intellectual development of students but also their access to desirable job opportunities, which can improve their quality of life. This paper focuses on school dropout in the Predict Students’ Dropout and Academic Success dataset. We use three resampling techniques for the imbalanced problem: Adaptive Synthetic, Support Vector Machine-Synthetic Minority Oversampling Technique, and Synthetic Minority Oversampling Technique+Edited Nearest Neighbor. We also compare the performance of the Random Forest, Support Vector Machine, and XGBoost classifiers between the default hyperparameter configuration and Bayesian configuration. The Synthetic Minority Oversampling Technique+Edited Nearest Neighbor technique obtained the best average performance using the Support Vector Machine classifier, achieving an accuracy of 93.55% and a precision of 94.11%.