Risk management is the commercial bank’s core business. As one of the most important risk, credit risks closely affect the development of banking. Under the circumstance of international financial crisis and global economic recession, it is crucial to do research on the quantification methods, techniques and models of the credit risk ranking.Support vector machine (SVM) is a convex optimization problem, which can used to analyze regression and classification, with the features such as small sample, nonlinear, high dimension. In the field of credit risk rating, the traditional parameter models have strict requirements of samples, which are not easy to meet. However, SVM can overcome the limits of traditional methods, so that researchers can apply SVM to rate credit risk for public companies, which apply for loan to commercial banks, avoiding personnel judgment with high experience risks.To begin with, this paper introduces logit regression and SVM, including the basic theory, principle, model and algorithm. The traditional logit model takes0.5as the boundary of classification, which exists misjudgment near the border, to solve this bug, this paper introduces SVM classification algorithm.Next, this paper applied financial data from the public company to verify the integrated algorithm. The experiments show that the integrated model has better classification effect and improved accuracy.In addition, this paper has the following innovative points:a) Using the principal component analysis to optimize variables and get new comprehensive variables, which covered more comprehensive information;b) Introducing quadratic forms into logit model, improved the prediction precision;c) Putting forward the integrated models, which combined of logit and SVM model;d) Presents the improved multi-classification algorithm.Finally, this paper summarizes the whole text briefly, and points out the further research directions. |