Research On Financial Crisis Early-warning Mdel Of Listed Companies In China A-stock Market

With the development of economic in China,the number of listed companies is increasing,while the number of loss-making companies is also rising.Therefore,it isimportant to establish an effective early warning model.In addition,If an effective early warning model of enterprise financial crisis can be established,enterprises will evaluate their financial situation in advance,and take effective measures to curb the deterioration of financial situation in time before the crisis erupts,which greatly reduce the possibility of bankruptcy,and also help investors and related enterprises to reduce losses.Therefore,it is of great significance to establish an effective early warning model of enterprise financial crisis for enterprises,investors and China's economic market.In order to establish the enterprise crisis early warning model of listed companies,this paper selects 72 indicators in 3372 financial data which contain 99 indicators.Then data filling,data induction and feature selection are carried out in turn.In the process of data filling,the common data filling methods are compared,and the random forest method with the smallest mean square error after filling is selected to fill the whole data.In the process of feature selection,17 feature variables were selected by using significance test,balanced random forest and Lasso method.Because the data in this paper is highly unbalanced,the classical algorithm,SMOTE algorithm and balanced random forest are used to resolve this problem.The early warning model is compared in recall rate,accuracy,macro average,F1-measure value and AUC value.Finally,it shows that the balanced random forest financial model is the best one.Therefore,the balanced random forest is selected as the enterprise financial early warning model,and the results of it are analyzed based on the actual situation.