<![CDATA[Student graduation times are often difficult to predict early, a major challenge facing institutions. Manual evaluations often fail to identify problematic students, leading to inaccurate graduation times that are detrimental to both students and institutions. This is crucial because study duration and timely graduation are important criteria in assessing institutional accreditation and quality. As an innovative solution, this study developed a graduation prediction model using the XGBoost and Random Forest algorithm, applying hyperparameter optimization techniques through Grid Search Cross Validation. The results showed that with default parameters, Random forest was superior to XGBoost. However, after hyperparameter tuning, XGBoost achieved better accuracy than Random Forest with a significant increase in accuracy, from 88.15% to 92.66% (precision 91.87%, recall 91.67%, and F1-score 91.38%). This confirms that appropriate hyperparameter tuning is a strategic key to maximizing the effectiveness of classification models. Thus, this model can be a tool for institutions to monitor and intervene early on in potential student delays.]]>
Copyrights © 2026
