<![CDATA[Loan classification is a critical component of credit risk management, as it categorizes loans based on risk levels and supports the financial stability of banks, where loan-related income represents a substantial share of assets. Effective classification aims to ensure secure asset allocation, minimize credit risk, and prevent potential repayment issues. This study enhances loan classification performance through two strategies: hyperparameter optimization of Decision Tree and Random Forest algorithms, and data balancing techniques to address class imbalance. Experimental results show that the Decision Tree achieves 89.21% accuracy with an F1-Score of 70.17%, while the Random Forest demonstrates higher performance, reaching 94.04% accuracy and an F1-Score of 79.75%. Random Oversampling reduces bias toward majority classes by improving model sensitivity, while hyperparameter tuning with GridSearchCV identifies optimal parameter settings, thereby strengthening predictive performance. The findings highlight that combining data balancing with hyperparameter optimization effectively improves accuracy and F1-Scores. These approaches are not limited to the algorithms tested but can also be applied to other classification methods, offering broader potential for enhancing credit risk prediction in banking.]]>
Copyrights © 2025
