Machine Learning Model for Small Bank Bankruptcy Prediction

Abstract

The bankruptcy of small banks can create serious disruptions to overall financial stability, especially in developing economies where the financial system is heavily supported by medium- and small-scale banking institutions. Unlike large commercial banks that often have diversified portfolios and greater access to capital buffers, small banks are typically more vulnerable to liquidity shocks, credit risks, and external market fluctuations. In this regard, advances in machine learning provide an opportunity to construct more robust prediction models that outperform traditional approaches, such as financial ratio analysis and logistic regression. This study explores various machine learning models, including Support Vector Machine, Random Forest, and Neural Network, while highlighting their strengths, weaknesses, and practical relevance in the context of small banks. By employing a Systematic Literature Review (SLR) method, the research systematically evaluates studies published and focus on bankruptcy prediction within the banking sector. The findings indicate that ensemble learning and deep learning models achieve higher predictive accuracy, though challenges in interpretability and transparency remain crucial for regulators and practitioners.