Hybrid Ensemble Approach for Higher Education’s Study Accomplishment Classification in Indonesia
Year: 2026
Type: article
Abstract: University student achievement plays a vital role in national development across scientific, economic, social, and other sectors. To enhance institutional efficiency, universities must ensure students graduate on time. Identifying students at risk of delayed graduation is therefore a critical task. While ensemble methods have been previously applied to predict educational outcomes, there remains room for improvement in their predictive effectiveness. This study introduces a hybrid ensemble model that combines Random Forest, XGBoost, and Stacking, using majority voting to generate final predictions. Unlike traditional bagging techniques, the proposed approach uses the same dataset across all models to maintain consistency and reduce variance. The model was evaluated using a dataset of 2,040 student records containing 22 features. It achieved 91.06% accuracy and 94.31% F1-score, outperforming individual ensemble models. Furthermore, a paired t-test confirmed that the performance improvement is statistically significant (p < 0.01). These results suggest that the proposed hybrid ensemble approach holds strong potential for addressing other classification problems in educational domains.
Authors: Anief Fauzan Rozi, Rahmat Gernowo, Adi Wibowo
Institution: Diponegoro University
Language: English
Subfield: Computer Science Applications
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: diamond
OpenAlex ID: W7128908528