An Empirical Study On Financial Early-warning Based On Principal Component Analysis

This thesis is about financial distress alert problem of listed companies. Currently, the problem has become to be one of the social focus issues, and how to establish an effective financial early-warning system for corporations to keep away the financial distress is the urgent assignment of our financial management. However, researches on estimation of financial distress alert are still in a beginning stage in our country, therefore, there are profound theoretic and realistic meanings in furthering researches.On the basis of forefathers'research results, this thesis introduce the theoretical model and application method of the financial distress alert. Then we regard the listed manufacturing companies that is treated specially(ST) because of unusual financial condition in A-Share market as the research object, and select 26 financial indexes as deterioration of liquidity and leverage, profitability, activity, growing ability, structure and cash flow as initial variables. Then we use them to do independent sample T test to pick up 16 financial indexes. At last, we use principal component analysis to get 8 ultimate variables. So we make the sample to two groups by random: learning sample and test sample. Then we use the Logistic regression analysis and BP neural network based on Principal Component Analysis to do the empirical study. SPSS,Eviews and Matlab are used to implement them. The results demonstrate the using the last three year's financial statement, this thesis's models could forecast the ST stock companies this year.