Predicting Financial Distress in a Turbulent World: a Comparative Machine Learning Analysis Across Nations

Abstract

This study evaluates the performance of six machine learning models in predicting financial distress, focusing on Indonesia and comparing with other nations. Using metrics like accuracy, AUC Macro, F1 Macro, F1 Weighted, and Log Loss, we find the Random Forest model with a Standard Scaler Wrapper performs best across most metrics, while LightGBM with MaxAbs Scaler is preferred for deployment due to its robustness and scalability. We analyze feature importance of identifying key factors influencing financial distress, such as investment growth, GDP growth, and economic uncertainty. Our findings highlight the critical role of machine learning in economic forecasting and policymaking, emphasizing the importance of digital optimization and AI-driven decision-making in addressing global financial stability.