<![CDATA[Objectively determining the Single Tuition Fee (UKT) level was essential for aligning educational costs with students’ socioeconomic conditions. This study compared Multiple Linear Regression (MLR) and Extreme Gradient Boosting (XGBoost) for predicting the final (Pleno) UKT level using data from 12,355 students, including socioeconomic, school-related, and administrative attributes. Data preprocessing involved cleaning, handling missing values and duplicates, transforming categorical variables, and splitting the dataset into training and testing sets (80:20). Models were evaluated under “operational” and “strict” scenarios using accuracy, precision, recall, F1-score, confusion matrices, and AUC. In the operational scenario, XGBoost achieved 93.42% training and 90.15% testing accuracy, outperforming MLR, which achieved 84.10% testing accuracy. Under the strict scenario, XGBoost accuracy decreased to 63.27–63.54%, compared with 49.74–49.86% for MLR. These findings demonstrated XGBoost’s greater ability to capture nonlinear relationships and complex attribute interactions. Permutation importance identified Coordinator- and Verifier-assigned UKT as the dominant predictors, while SHAP analysis identified per capita income as the most influential socioeconomic factor. Overall, XGBoost showed superior predictive performance but should be used as a decision-support tool rather than as the sole determinant of student tuition fees.]]>
Copyrights © 2026
