Prediction of Financial Crisis Based on Machine Learning

Abstract

The financial crisis is an inevitable catastrophic event in the operation of the entire capital market. And it may cause significant losses to the entire market. So as individuals, if they can predict and respond in advance, it will reduce a lot of losses and make the company's life cycle longer. Here we used the data on credit defaults with a total sample of 1,000 samples containing Germany's credit default records and some basic personal information. Logistic regression, random forest and Xgboost were applied to discover useful information behind these data. The results showed that the machine learning method fitted the data relatively well, and the accuracy of Xgboost has reached about 80%. Existing checking account and foreign worker were two most important indicators to help predict financial crisis. In this way, both companies and the country could reduce their losses, so that they can spend the time of the financial crisis more smoothly and promote social prosperity.