Financial crisis predictions based on biorthogonal wavelet hybrid kernel support vector machine

Abstract

Financial crisis prediction is generally researched by means of the kernel principal component analysis (KPCA) and the support vector machine (SVM) model. However, kernel functions used in these methods are basically single. In fact, hybrid kernel functions are superior to the component kernel functions in dealing with non-linear issues, for they can make full use of the feature mapping abilities of different kernel functions. Based on biorthogonal wavelet kernel functions CDF9/7 and linear kernel functions, a new type of biorthogonal wavelet hybrid kernel functions are constructed. Besides, KPCA-SVM models based on hybrid kernel functions are also proposed for financial crisis prediction. The empirical research on the listed companies in China's securities market is conducted at last. The results show that the new type of biorthogonal wavelet hybrid kernel function can improve the feature extraction of KPCA and enhance the prediction accuracy of SVM model, and then the accuracy of the financial crisis prediction can be improved.