Principal Component And The Combination Of Support Vector Machines Listed Companies' Financial Early Warning Model

With the deepening of economic reform in China, more and more companies will have to be faced with fierce competition. In Chinese developing but not mature security market, whether the listed companies are healthy or not, will affect related interests of investors , creditors and itself, and it also block healthy development of Chinese Securities Market. Therefore, if we can establish a practical and effective pre-warning model on financial crisis to generalize the researchful effective results, it will be more significant to guide the management and risk prevention of the enterprises.Recently lots of methods are applied on building pre-warning model on financial crisis, including traditional methods, for example, Bayesian method, distance distinguish, Fishier distinguish, PCA, etc; Recent methods such as fuzzy classification, coarse classification and networks, as well as support vector machine.First, this Paper points out financial implications, functions and features for the necessary analysis, and reviewed the researches on the financial distress prediction model at home and abroad, then forms a new combination forecast method, using combination forecast theory to combine the PCA with SVM, to build pre-warning model on financial crisis of IT industry with 64 listed companies of IT (Information Technology) as the study objects. After doing the experiment, the result shows that this combination forecast model has the advantages of both tradition statistics model and SVM, and the accurate ratio of trained samples is 100%, tested samples is 85%.In order to test the validity of financial distress prediction model, we use the same samples and financial data to establish a logistic regression model, through comparing the accuracy of these two forecast models, we find that the combination model has higher accuracy.