Study Of Application Of Credit Rating Model To Appraisal Of Credit Risk

It is important, both theoretically and practically, for banks and investment companies to evaluate loan borrowers'credit risk. The main purpose of this thesis is to find out how to quantify credit risk. Linear-Discriminant analysis, Probabilistic Neural Network, Back Propagation Neural Network and Support Vector Machine are all adopted to develop credit rating models. To find the best and practical method, a quantified analysis and experimental study are emphasized.Firstly, in this thesis the concept of credit rating is presented and problems existing in credit risk appraise field are summarized. Secondly the advantages and disadvantages of existing methods are analyzed and the adaptability of existing methods in China is presented. Some popular techniques for developing credit rating models, such as Linear-Discriminant analysis, Probabilistic Neural Network, Back Propagation Neural Network are introduced. In addition, Support Vector Machine for credit rating is presented. Lastly, four techniques were applied to measurement of credit rating for listed companies in China, and the classification accuracy is also compared. More over, Support Vector Machine is also proposed for classification of small samples.To sum up, finance management, econometrics and information technique are synthetically taken into account, and various measures are applied to building credit rating model in this thesis. The measurement manifests that the four models have a good ability of classification.