The Research On Financial Crisis Early-warning Model Of Listed Companies Based On Predicted Value

Enterprises are threated by financial crisis in the fierce stockmarket now. If enterprises can't take precaution against the financial crisis in time, it will get into trouble. But enterprises are not powerless in the face of such a predicament, the financial crisis can be predicted in advance. Thus, essential analysis based on setting up the financial crisis early warning index system which is disrelated and covered amount of information, and building up a financial crisis early warning model, has become the important aspect of stabilifying the development of stockmarket, national economics, and society.The content of chapter one analyses the study of domestic and foreign references, and points out some deficiencies, such as samples ignoring the difference of industry diversity, warning index excluding the cash flowing, discrimination analysis without regard to the hypothesis, and financial crisis warning model being not dynamic. The second part analyses the meaning, characteristics and cause of the enterprises' financial crisis. The third part firstly chooses 110 A-listed companies from Stock Exchange of Shanghai and Shenzhen as research samples (55 companies first with and 55 without financial crisis). And 6 warning indexes are eliminated by Independent-Sample T Tests and Nonparametric Tests. And then it sets up listed companies' financial crisis warning index system including 10 principal component factors using the method of principal component analysis. The fourth part builds up the Canonical Discriminant Function by 10 principal component factors as discriminating independent variables based on Fisher Discrimination Rules. According to the 2000-2003 financial ratios of 40 testing samples, the fifth part puts out the validity result of warning model through the forecasted financial ratios of 2004.The main characteristic and innovation of this paper lies in four aspects. Firstly, the index system is more integrate for including cash flow ratios. Secondly, there are twice-filters for 27 chosen original financial ratios. The final 10 extractive factors involve the 89.746 per cent of all the original information, by principal component analysis. Meanwhile, it avoids the relativity between all the factors and improves the veracity of later prediction. Thirdly, this paper combines the principle component analysis and discrimination analysis, avoiding the lack of effective information. Finally, this paper adopts three different methods to test the validity of financial crisis warning model, that is self-validated, cross-validated, and forecasting-validated. The separate correct rate is up to 90 per cent, 88.2 per cent, and 80 per cent. The most important is that the forecasting-validated method which adopts GM(1,1) Grey Forecasting has got the aim of long-term warning.