| The security market is the important base of modern economy, the performance of security market is closely related to the real economy。As the main part of security market, the performance of the listed company influence the running status of security market。In order to protect the profit of investor and detect the sign of enterprise financial risk symptoms, it is vital important to establish an early warning mechanism of financial risk, which is maturity。If the financial data is open and transparent, it will avail to establish an early warning mechanism of finance, therefore, establish an applicable and precise financial early warning system to predict the enterprise crisis has become a long standing theory research hotpot。Now statistical method and data mining algorithm become the trend of theoretical research of financial risk evaluation, every coin has its two, both methods have its advantage。As the traditional statistical method, the forecast result of discrimination analysis and regression analysis play well, however, the statistical hypothesis is rigorous, data could not meet the hypothesis。Data mining algorithm includes neural networK、decision tree、SVM (support vector machine) and so on, these methods have few rigorous hypothesis constraint and powerful study ability, which have broader applied range。Both statistical methods and data mining have its advantage, so we can not abandon any of those methods and adopt another method only, as when you abandon other method, you might ignore some information that matter。The emergence of combination forecast theory resolve this puzzle problem, combination forecast adopt two or more forecast methods to the same problem。It could be either combination of several quantitative method or combination of several qualitative method, but combination of quantitative method and qualitative method is need in practice。The purpose of combination is to take advantage of the information of every method, in order to improve prediction accuracy。... |
PDF Full Text Request