Research on Asset Quality Detection Based on Hybrid Biorthogonal Wavelet OCSVM Model

Abstract

Asset quality is the foundation of enterprise survival and development we choose one-class support vector machine (OCSVM) is chosen to deal with asset quality abnormal detection for it pays great roles to acquire the abnormal data. As well as flexible to be constructed, Biorthogonal wavelet consists linear-phase nature and high vanishing moment, therefore, corresponding wavelet kernel functions are respectively constructed relying on Bior (2, 2) and Bior (3, 9) wavelet. On the basis of which new hybrid kernel is proposed and then introduced to OCSVM to innovate the model. In addition, it is applied into kernel principal component analysis (KPCA) method to realize a high dimension space mapping and enforce dimension reduction. Empirical research on A-share listed manufacturing enterprises at last is conducted and the result tells that the model we come up with is greatly improved on the recognition rate of abnormal samples when compared with other method.