Predicting Financial Crises: A Machine Learning Approach

Abstract

This paper examines the ability of econometric and machine learning techniques to predict the financial crises. Our findings show that the traditional econometric models underperform and unable to predict financial crises in out of sample. The machine learning algorithms such as extree, random forest of ensemble methods improves the accuracy. Moreover, prediction results confirm that the importance of external vulnerable indicators play crucial role in predicting financial crises using Shapley values.