Research On Financial Early Warning Of Listed Companies Based On Stochastic Forest And XGBoost

With the rapid development of economy and the increasingly fierce market competition,enterprises are facing more and more risks and crises,and many enterprises are facing bankruptcy crisis due to financial problems.In order to alleviate this situation,it is necessary to warn the financial situation of enterprises so that managers can find the crisis signals and take effective measures to reduce the possibility of crisis outbreak,and maintain the normal operation of enterprises.Establishing financial crisis warning model has become an important direction of theoretical and practical research.This paper takes 196 A-share listed companies in Shanghai and Shenzhen Stock Exchanges from 2017 to 2018 as research samples.According to the selection criteria of sample data,50 financial early warning indicators are selected from eight aspects: the company's T-3-year operating ability,development ability,profitability,solvency ability,cash flow index,per share index,ratio structure index and relative value index.This paper mainly studies the screening of financial crisis early warning indicators and the construction of financial early warning model.(1)Constructing the second-level financial early warning index system.Before building the model,all the financial indicators are pre-processed by R language,and then the financial early warning indicators of the entry-exit model are screened,including statistical test of sample data and screening indicators by random forest principle.Using the results of K-S test to test the indicators that obey normal distribution,and Mann-Whitney U nonparametric test to test the indicators that do not obey normal distribution,to screen out the indicators that can clearly distinguish ST listed companies from non-ST listed companies.On this basis,using the principle of stochastic forest,the indicators with high information content and low correlation are extracted as secondary financial early warning indicators.Standard system is used to construct early warning model.The empirical results show that 39 indicators that can distinguish ST sample companies from non-ST sample companyies are screened out by statistical test.After further reducing the dimension of indicators by using random forest,13 financial indicators,such as P/E ratio,total asset growth rate,net profit cash content,cash content of operating income and net profit rate of total assets,are obtained to constitute the second-level financial early warning index system.(2)Establishing a financial crisis early warning model.Based on the financial index system,the financial early warning model is constructed by using XGBoost principle,and the data of training set and test set are tested and predicted.The results show that the prediction accuracy of XGBoost early warnin g model based on random forest is 89.17%.By comparing the model with the classical financial early warning model,we can find that the model can better warn whether the company will fall into crisis and meet the needs of financial early warning.Finally,it points out the prospects of the research and gives further research space for financial early warning.