<![CDATA[Student retention is a critical indicator in evaluating the quality of higher education institutions. High dropout rates pose significant challenges, including at Al-Khairiyah University in Cilegon, Banten. This study develops a predictive model for student retention using two machine learning approaches: Recurrent Neural Network (RNN) and Support Vector Machine (SVM), while identifying the most influential factors. The dataset comprises 3371 records from 2021-2024, including academic variables (GPA, semester grades 1-8, attendance) and non-academic variables (organizational activity, competition achievements, parental income, admission pathway, and study system). Data was split into 80% training and 20% testing sets. Results show that the RNN model demonstrates superior performance with 93.5% accuracy, 99.7% precision, 89.3% recall, 94.2% F1-score, and 0.967 AUC, while SVM achieved 85.5% accuracy, 89.8% precision, 85.3% recall, 87.5% F1-score, and 0.912 AUC. Feature importance analysis reveals that Total GPA and first-semester grades (IPS.1) are the dominant factors influencing student retention, while non-academic factors have relatively small contributions. This research provides practical contributions through an Early Warning System framework that can be implemented by universities to detect at-risk students early, enabling proactive academic interventions.]]>
Copyrights © 2026
