An Empirical Prediction Research Of The Financial Distress Enterprises

With the advancement of global economic integration,how to survive and develop in such a complicated environment is not only the problem that investors,creditors and enterprise managers consider but also the government?s or a nation?s top priority.In such an unstable marketing environment,Some listed companies are marked by ST,what to be worse,Leading to the occurrence of bankruptcy.Therefore,it is very important to take relevant measures to protect the interests of investors,creditors and enterprises before the financial distress occurs.After reviewing the related literature,Based on the analysis of the definition of financial distress and financial distress influencing factors,this paper analyzes and summarizes the financial distress model used by the previous scholars.This paper chooses Three Mathematical Methods consist of Logistics model,principal component analysis model,and BP Neural Network Model.This paper selects 50 ST manufacturing enterprises in 2016 and 100 non-ST enterprises as a study sample to study the financial indicators and the ownership structure of enterprises,and synthesizes all the factors which may affect the financial distress,including cash flow ability,Solvency,development,profitability,operating capacity and equity structure of the Company.34 financial warning indicators were selected in t-2 year's financial indicators and the normal distribution test was carried out based on these financial indicators.Two independent samples T test were performed on the indexes obeying the normal distribution,and Two Independent Samples Nonparametric Tests were performed on the indexes not obeying the normal distribution.Based on the results of two independent samples T test and two independent nonparametric tests,8 financial indicators were not significant.Therefore,8 financial indicators are removed.Then,the indexes of multicollinearity in each financial capability concluding liquidity,solvency,development,profitability and operating capacity are eliminated,and the related indicators in each financial capability entered in the following empirical models.In the multicollinearity test,a total of 10 indicators were eliminated.At last,the remaining 16 financial indicators are provided in the Logistics model,principal component analysis model,and BP neural network model.All in all,In three kinds of warning model methods,The BP neural network has the advantages of self-learning,strong adaptability and intelligence,and the prediction accuracy is 90.7%,but it takes up the most time and other resources.Logistics model analysis requires that multiple variables do not have multicollinearity.If there is multicollinearity between variables,the model will lose its meaning.In the empirical study,the fitness of training samples is very good;The prediction accuracy of the test sample is 86.7% only lower than the BP neural network method.Principal component analysis uses a small number of six variables instead of the original total of 16 variables to solve not only the problem of too many variables and multiple collinearity of the original data but also retain most of the information;As the 6principal components in the principal component analysis reflect 80% of the information of the total original variables,The accuracy of the final model is only 80% in the lack of information integrity above,which is lower than the other two methods.