The Financial Distress Based On Machine Learning

The normal operation of the company is the basic element for maintaining social stability,and it is also an important prerequisite for ensuring the development of the country.However,the company's daily operations are not stable.Often,the company's operations are unstable or loss-making due to internal management problems,decision errors,or external macro factors,and it will even cause the company to fall into financial difficulties.In the crisis,this makes it especially important to study the financial distress of companies.In recent years,with the rise of the field of machine learning,it is of great practical significance to use machine learning models to study the company's financial distress.Based on macro and micro factors,this article conducts an empirical analysis of the company's financial distress from the following three aspects:First,the relationship between external macroeconomic factors and the company's financial distress was discussed.The China Economic Policy Uncertainty Index(CEPU Index)compiled by Shangqin Lu and Yan Huang was selected to use the DCC-MIDAS model to study economic policy uncertainty and the ST sector Dynamic correlation between the two.The study found that the volatility sequence between the CEPU index and the ST sector index has a positive impact on the whole,and it has been found that changes in the ST sector index volatility are affected by domestic events,while CEPU index volatility changes are more affected by international events;The dynamic correlation between the CEPU index and the ST sector index shows a positive correlation on the whole and fluctuates greatly.In addition,the correlation between frequent positive and negative changes is affected by domestic and foreign events.Secondly,the economic policy uncertainty index(EPU index)in China compiled by Baker et al.was used to establish Logit model to explore how the economic policy uncertainty and corporate financial indicators affect the financial distress of listed companies.The study found that: when the uncertainty of economic policy increases,the probability of merger financial breakthroughs of merged listed companies;Equity ratio and other four financial indicators have an impact on the company to reduce the probability of financial distress;The lag EPU index has the same effect on the financial distress as the EPU index over the same period.Further research on the impact of economic policy uncertainty on the financial distress of listed companies with different property rights,and found that non-stateowned listed companies are more affected by economic policy uncertainty.Finally,based on the data of listed companies and the empirical results above,we use machine learning models to discuss the early warning of the company's financial distress,and predict the financial distress of A-share listed companies based on four machine learning models such as the random forest model.The study found that the use of machine learning technology can significantly improve the prediction accuracy,of which the random forest model has the best prediction accuracy.In further research,we discussed the impact of macro factors on machine learning models to study the company's financial distress forecasting problem by adding macro indicators.As a result,it was found that feature vectors containing macro factors can make machine learning models generally increase the accuracy of prediction.