Machine Learning and Parameter Optimization for Banking Stability Prediction and Determinants Identification in ASEAN

ASEAN Bank Financial Stability Parameter Optimization Machine Learning Predicting

Authors

  • Pham Thuy Tu Ho Chi Minh Universiy of Banking, 56 Hoang Dieu 2, Thu Duc, Ho Chi Minh City, Viet Nam
  • Dao Le Kieu Oanh Ho Chi Minh Universiy of Banking, 56 Hoang Dieu 2, Thu Duc, Ho Chi Minh City, Viet Nam
  • Do Doan Trang
    doantrang.bolt@bdu.edu.vn
    Binh Duong Universiy, 504 Binh Duong Boulevard, Thiep Thanh Ward, Thu Dau Mot City, Binh Duong Province, Viet Nam
Vol. 9 No. 3 (2025): June
Research Articles

Downloads

This study leverages machine learning and advanced variable selection techniques to enhance the prediction of the Bank Financial Stability Index (Z-score) in emerging ASEAN markets. Utilizing a comprehensive secondary dataset comprising macroeconomic and bank-specific indicators from 61 commercial banks across Indonesia, Malaysia, the Philippines, Singapore, Thailand, and Vietnam (2010–2023), we systematically evaluate the predictive power of multiple machine learning models. A rigorous cross-validation framework is employed to optimize forecasting accuracy, integrating Linear Regression, Random Forest, K-Neighbors, Decision Tree, Gradient Boosting, AdaBoost, Support Vector Regression, and XGBoost with Lasso, Ridge, and Elastic Net regularization. Empirical results reveal that key drivers of financial stability include equity capital, financial leverage, return on equity, GDP growth, inflation, technological advancements, and systemic shocks like the COVID-19 pandemic. Notably, the Ridge-optimized XGBRegressor model achieves the highest predictive accuracy (~89%), demonstrating the efficacy of hybrid machine learning approaches in financial stability forecasting. These findings offer crucial insights for policymakers and regulators, facilitating data-driven strategies to strengthen banking resilience and mitigate systemic risks in volatile economic environments.

Jel Classifier: C45, C52, C55, G21, G32.