An Empirical Study On Financial Crisis Early Warning Of Listed Companies Based On Different Feature Selection

With the establishment of the market economy system and the rapid and steady development of the economy,the listed enterprises gradually become an important promoter of the development of the national economy and play an increasingly important role in the national economy.But the finance is the lifeblood of the enterprise development,the financial condition appears the question not only can cause the enterprise own stable operation,the entire correlation market normal movement will suffer the significant influence.Therefore,it is of great significance to monitor and forecast the financial status of listed companies.In this paper,the special treatment of listed companies by the Shanghai and Shenzhen stock markets is taken as a sign of financial crisis,and the first special treatment of listed companies in 2017-2019 is taken as a sample of financial crisis,according to the ratio of 1:3,select the listed companies that have never been specially treated as a normal financial sample.36 indicators including 8 aspects of economic value added were selected from both financial and non-financial aspects,and the final 28 indicators were determined after preliminary screening of the indicators by using statistical methods,then,28 indexes are selected by two different feature selection methods,and 28 final indexes and the indexes obtained by two different feature selection methods are taken as the input variables of the early warning model,the influence of different feature selection on early warning model is analyzed.The empirical research shows that:(1)the variable after feature selection can eliminate the unnecessary information in the original index,which makes the early warning model have better prediction precision and stability.(2)the classification accuracy and validity of early warning model based on RF?RFE is better than PCA.(3)the random forest algorithm and the XGBoost algorithm are more suitable for constructing the financial crisis early-warning model.(4)the introduction of Eva Index is effective for the construction of financial crisis early warning model,and the four indexes of eps,total assets net profit rate,Yield valve,and net assets EVA rate are the important influencing factors for the occurrence of financial crisis,enterprises should pay attention to the production and operation activities of the four indicators involved in the area of financial crisis prevention.