Loan Default Forecasting Using Machine Learning Algorithm in Indonesian Regional Bank

In an effort to address the issues of growing credit risk, this research analyzes the application of machine learning algorithms for loan default prediction in regional banks in Indonesia. This study evaluates the efficacy of four primary algorithms - Decision Tree, Random Forest, Gradient Boosting, and Naïve Bayes - by analyzing a two-year historical loan dataset. In-depth study indicated that Naïve Bayes is the best algorithm, attaining 99.96% accuracy, 98.45% precision, and 99.63% recall, with a perfect Area Under the Curve (AUC) score of 1,000. These results represent the improved capability of Naïve Bayes in identifying probable defaults with high precision and accuracy, greatly minimizing the probability of false positives. The findings recommend the application of Naïve Bayes as a significant method in local bank credit risk management, providing an effective solution for early default identification and increased financial stability. This study adds to existing literature by presenting actual proof of the efficacy of machine learning technology in predicting credit risk. The suggestions offered can aid regional banks in enhancing their risk management approaches.