| In April 2007, the Chinese stock market facing unprecedented bull market, so that people won't be a heady, however, followed by the US subprime mortgage crisis broke out, the company's financial crisis, the negative messages can also make investors lose huge, pay social costs are enormous. In the financial crisis before, if you are able to anticipate the possibility of the financial crisis of the United Nations as well as the various stakeholders, to take the necessary measures in order to prevent greater financial crisis broke out. Therefore, the establishment of an effective financial mechanism for warning die Government circles, of great significance.This article is listed as a result of the financial situation of the exception is special treatment (ST) as the enterprise into a financial crisis, is a collection of sample data 336 (224 non-ST companies, 112 ST companies) of the company, this article is paired with a 2: 1 Select paired two sample overcome at predictive modeling because paired two sample too low due to bankruptcy, the chances of getting higher than the actual odds of bankruptcy. This article lists 31 financial targets, on this basis, by the extraction of principal components analysis model variables, and eventually obtain 10 principal component variables are used to predict the financial crisis. Then use Logistic model respectively, BP neural network and support vector machines three methods are warning model of building and testing, and on different models of comparison; the results found that the integrated Logistic model forecast accuracy rate of 82.5%. Neural network model forecast accuracy rate of 85%. Support vector machine model forecast accuracy rate of 86.67%. Support vector machine model works best, neural network model par support vector machine model, and the first two Logistic model forecast than the worst effects. Finally the relevant conclusions of the study and further elaborate on the limitations of this study and future research directions. |
PDF Full Text Request