Study On Credit Risk Assessing Model In Commercial Bank

Credit risk facing the banking sector is important. Research of bank credit risk is a basic work. Since 2007 accession to the WTO, foreign banks further erode the domestic market. Chinese commercial banks face increasingly intense competition. The New Basel Capital Accord is implemented to domestic commercial banks have brought new challenges. In such circumstances, many domestic and foreign researchers in this area had extensive and in-depth research, design a lot adapted to the specific environment of the credit risk assessment model. Neural Network and SVM are simulations of intelligence of the people. Both of them can establish a non-linear model after a given system's input/output data. They have been widely used in control, decision-making, expert system, and model identification fields during the past ten years. This paper studies the advantages and disadvantages of SVM and Neutral Network. At the same time these theories are applied to the bank's credit rating, which is a great step from theory to practice. And this paper consist of the following several parts.(1)After the analyzing the theory of neural network,credit risk detection approach based on the artificial neural network is presented. It uses BP net and RBF net to designed, whose training algorithm is improved to get faster convergence and more accurate direction. This paper designed and programmed an assess credit risk rating model on a commercial bank based on MATLAB system.(2) The SVM algorithm uses two kind of kernel function: Radial Basis kernel, Polynomial kernel to classify the credit risk data. These classifications are all multi-class classification using in ctedit risk. The assessment result is proved right by our test of models. As a result, the Radial Basis kernel SVM model used to assess credit risk in a commercial bank is better than others. Finally, this paper gives ananlysis and conclusions.