<![CDATA[The increase in non-performing loans in the SME sector represents a major challenge for the banking industry, as it directly affects financial stability and profitability. Conventional credit risk assessment approaches are considered insufficient in detecting potential defaults at an early stage. This study aims to apply a data-driven approach to determine effective strategies for managing non-performing loans in the SME banking sector. The research utilizes historical SME credit data from Bank X Regional 11 and applies a Random Forest machine learning algorithm to predict credit collection. The analyzed variables include payment behavior, credit structure, and the financial condition of debtors. The results indicate that the last payment date is the most influential variable in predicting credit risk, followed by loan tenor, loan realization timing, interest rate, and savings balance. The Random Forest model demonstrates high accuracy and stability in credit risk classification. Based on these findings, non-performing loan management strategies are formulated using the G-STIC framework integrated with the 3-C approach (Character, Capacity, Capital), emphasizing early warning systems, credit structure adjustments, and debtor-based risk control. This study provides practical insights for banks in enhancing objectives and sustainable SME credit risk management.]]>
Copyrights © 2026
