<![CDATA[Predicting the Single Tuition Fee (UKT) in economically vulnerable areas remains challenging due to diverse socioeconomic conditions and the limited capacity of existing models to capture interactions among contributing factors. This study proposes a hybrid framework that integrates segmentation, pattern discovery, and classification to improve UKT prediction accuracy in the Southwest Aceh region. Data from 452 student records were analyzed through three main stages: cluster-based socioeconomic segmentation, association pattern discovery to reveal dominant factor relationships, and the integration of these hybrid features into a Random Forest model. To ensure model stability and prevent data leakage, the framework was evaluated using a multilevel 5-fold cross-validation strategy. The developed model achieved 93.4% accuracy, demonstrating a significant improvement over baseline demographic models while identifying key socioeconomic determinants of UKT assignment. The proposed framework demonstrates how hybrid learning strategies can improve fairness and transparency in educational financial decision systems. These findings highlight the potential of the hybrid framework to support fairer and more data-driven tuition policies, offering a robust evaluation design for educational finance modeling in Indonesia.]]>
Copyrights © 2026
