| In China's stock market listed companies was treated specially by CSRC because of their financial problem every year, which not only can make their own losses, also bring economic losses and negative impacts to stakeholders, and even cause the bad effect on market environment. In view of the reason that China's securities market is still in a weak active phase, information asymmetry is a common phenomenon, therefore, the financial crisis warning model for the prediction and prevention of listed companies' financial crisis is necessary.Firstly, this thesis has combed the financial crisis literature systematically, respectively, from the definition of the financial crisis,financial crisis early warning indicators selection, financial crisis warning model and the sample selection of financial crisis to summarize a conclusion; then, extracted financial indicators used frequently, from the financial crisis research literature, building technical indicators to measure the financial crisis and to standardize data processing; Finally,nearest neighbor domain method is adopted to establish the financial crisis warning model. By collecting financia l data of all listed companies in the stock market, comparing these data with each other by means of the SVM model, analyzing and forecasting accuracy of two models and two types of error rate, the results show: The second type of error of SVMmodel was zero, while the Type I error rate table is higher, which shows that the early-warning effect of this method is excessive. Nearest neighbor method was relatively stable: first class and the second class error rates are relatively moderate, which is more conducive to asset selection and risk aversion of investors.K-nearest neighbor algorithm based on the financial crisis early-warning model developed in this study has high practical value,which can provide valuable reference for stakeholders interests of listed companies, especially those with a good guide for investors decisions. |
PDF Full Text Request