PREDIKSI WAKTU KELULUSAN MAHASISWA BERDASARKAN FAKTOR AKADEMIK DAN DEMOGRAFIS MENGGUNAKAN RANDOM FOREST DAN XGBOOST

Accurate graduation time is an important measure to illuminate how well the higher education system functions. Data from 10,000 students was used, including GPA, credits, age, gender, place of residence, employment status, economic status, and scholarship acceptance. Class imbalance in the data is addressed through the CRISP-DM and SMOTE methods. The evaluation results show that both algorithms have the capability to predict permit status with high accuracy; Random Forest achieved an accuracy of 91.95% and XGBoost 91.85%. Based on the precision, recall, and F1 score, both models demonstrate very good and balanced performance, with Random Forest being slightly superior in result stability. Therefore, Random Forest is recommended as the best model for graduation prediction. This research is expected to help colleges identify students who may graduate late to provide timely interventions.