Abstract: Bank liquidation is a process of resolving the obligations of banks whose licenses have been revoked and plays a crucial role in maintaining financial system stability. The duration of Rural Bank liquidation varies depending on financial conditions and asset recovery performance, requiring robust analytical methods for accurate classification. This study aims to classify the liquidation duration of Rural Banks using the Random Forest algorithm with class imbalance handling through the Synthetic Minority Oversampling Technique (SMOTE) and hyperparameter optimization using GridSearchCV. The dataset consists of secondary data of Rural Bank liquidations handled by the Indonesia Deposit Insurance Corporation during the period 2006–2024, comprising 120 observations and 22 financial variables. Data preprocessing includes median imputation for missing values, winsorizing for outlier treatment, and data standardization. The results indicate that the best-performing model is Random Forest with SMOTE and hyperparameter tuning, achieving an accuracy of 83.3%, precision of 75.0%, recall of 93.8%, and F1-score of 83.3% for the problematic class. Feature importance and SHapley Additive exPlanations (SHAP) analysis reveal that Asset Liquidation Estimate, Total Gross Assets, and Guarantee Claim Value are the most influential variables. These findings provide important implications for data driven decision making in bank resolution policies.
Maulana et al. (Mon,) studied this question.