Predicting Models of Financial Crises

Abstract

This study used data from 1870 to 2008 to test the role of logistic regression models in predicting financial crises. This article first dealt with missing values in the data and removed time trends. Then, logistic regression models were used to predict and explore the occurrence of financial crises. During the exploration process, this article adopted principal component analysis and the method of eliminating multicollinearity interference for model optimization, and ultimately found that the area under the curve (AUC) in logistic regression was the highest, at 0.86. This indicates that logistic regression models provide more reliable predictions for financial crises. Based on the results, we further speculate that the total asset value and GDP are very important factors determining whether a financial crisis will occur. While there is room for improvement, the conclusion for this study still provides valuable insights into the connection between financial crises and specific banking factors.