<![CDATA[The Surabaya City Government launched the ‘Dandan Omah’ Social Rehabilitation Program for Uninhabitable Houses (RTLH) in 2022. However, beneficiary selection for the 2022–2024 period was conducted manually, which may have led to subjective and less targeted decisions. This study proposed a data-driven decision support approach by comparing Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN) across 2, 4, and 5 classification tasks. A dataset of 1700 records comprising 11 features was evaluated using an 80:20 train–test split and 10-fold cross-validation. RF consistently achieved the best performance across all scenarios while SVM demonstrated competitive baseline performance. After tuning, DT surpassed SVM on 5-class tasks and matched RF on 2-class tasks, making optimized DT a viable interpretable alternative when decision accountability was essential. KNN consistently underperformed due to the curse of dimensionality. Overall performance declined as classification complexity increased, attributed to variability in field assessment standards. These findings suggested that RF and optimized DT could support more standardized and objective RTLH beneficiary selection. Future studies employing RF were recommended to identify influential input features to optimize model performance while maintaining explainability. Re-standardization of field assessment protocols was also recommended to improve data quality.]]>
Copyrights © 2026
