The Research On Financial Distress Dynamic Prediction Of Listed Corporations With Support Vector Machine

To establish an effective suitable financial warning system has the very important realistic meanings for the listed companies and its benefit correlations, for it can send a dangerous signal when the company faces financial risk. So the Purpose of this paper is to find warning indicators that can remarkably distinct the financial crisis companies and the non-financial crisis companies through analyzing financial indicators of the listed companies and establish warning model on financial crisis that is suitable for the listed companies of our country, provide the making-Policy basis and the advice for related respects.This article regards the listed company that is treated specially (ST) in 2007 because of"unusual financial condition in A-share market as the research object. We select 94 companies training samples (composed of 47 ST companies and 47 non-ST companies). This article studies the data characteristics of financial indicators of sample companies. In addition, assessing system, which consists of 44 indexes and could be extracted to 13 factors with definite economic meanings by factor analysis.Then support vector machines and fuzzy integral are introduced. And the essay particular dissertate structure and calculation principle of support vector machines. Support vector machines ensemble has been proposed to improve classification performance recently. However, currently used fusion strategies do not evaluate the importance degree of the output of individual component SVM when combining the component predictions to the final decision. A SVM ensemble method based on fuzzy integral is presented in this paper to deal with this problem. This method aggregates the outputs of separate component SVM with importance of each component SVM, which is subjectively assigned as the nature of fuzzy logic.The result of basing on single dynamic SVM, fuzzy integral ensemble static SVM are compared in the essay. Fuzzy integral ensemble SVM is better than two of others. By the figures and tables that the result of basing on fuzzy integral support vector machines is satisfaction, it demonstrates the proposed method is stable, highly accurate, and feasible. It is useful for providing a sound financial distress prediction system for companies.