<![CDATA[Non-Performing Loans (NPLs) represent one of the most critical indicators in assessing the stability and financial health of banking institutions. An increasing NPL ratio may negatively affect profitability, liquidity, and overall banking performance, making accurate prediction models essential for effective credit risk management. This study proposes a predictive framework for estimating Non-Performing Loan levels using the Extreme Gradient Boosting (XGBoost) algorithm combined with Explainable Data Mining techniques. The dataset consists of historical banking and financial indicators that influence loan repayment behavior. Data preprocessing stages include data cleaning, feature selection, normalization, and handling of missing values to improve model performance. The XGBoost model was employed due to its ability to handle complex nonlinear relationships and high-dimensional data efficiently. To enhance model transparency and interpretability, Explainable Data Mining techniques based on SHapley Additive exPlanations (SHAP) were applied to identify the contribution and importance of individual features affecting NPL predictions. Experimental results demonstrate that the proposed XGBoost model achieves high predictive performance, outperforming conventional machine learning approaches in terms of accuracy, precision, recall, F1-score, and Area Under the Curve (AUC). Furthermore, the explainability analysis reveals the most influential factors contributing to NPL levels, providing valuable insights for financial institutions and policymakers in developing proactive risk mitigation strategies. The findings indicate that integrating XGBoost with Explainable Data Mining not only improves prediction accuracy but also enhances the interpretability and trustworthiness of credit risk assessment models. This approach can support data-driven decision-making processes and strengthen the sustainability of banking operations in increasingly complex financial environments.]]>
Copyrights © 2026
