Study On Credit Risk Classification Rating By Mathematical Model

Credit risk is the loss possibility suffered by the financial institution, the investor and the other sides in fiance because the loaner, bond publisher or one side in the fiance dose not perform the agreement. In a more restricted sense, credit risk is loss possibility suffered by financial institution since the borrower has not the ability or does not want to perform the agreement of repay the capital and interest, i.e. it is a behavior breaking a contract.Under the enviroment of modern market economy, the credit risk of enterprises is the basic activities of the financial system and financial institutions. Researching on how to improve the enterprises’credit risk management and then findinng proper credit risk identification system being suitable for chinese enviroment has great significance to promote and deepen the theoretical investigation on corporate credit risk, especially to enhance the ability resisting risks of commercial banks in our country. On the basis of the current problems existing in credit risk rating of corporation and the analysis of relevant achievement domestic and abroad, this thesis formulates three mathematical models modelling the credit risk rating of corporation. The K-Means cluster, the Self-organizing competitive neural network and the fizzy neural network are some of the important method used in the present thesis.The main constent of the thesis is to construct mathematical model modelling the rating of corporate credit risk identification and consists of five parts:introduction, the basic mathematical theory, building index system, empirical analysis of the shortage and importment of the model and the better of system. Firstly, the background and significance, the basic concepts of credit risk, the main methods and the development of credit risk are described in the introduction, respectively. In the second part, the basic mathematical methods, such as the cluster analysis, the fuzzy theory, the neural networks and the fuzzy neural network, are summarized systematically. Then, a comprehensive index system reflecting the financial situation is estabilished according to the general logic in corporate credit risk. In the fourth part,48financial datum for different enterprises in the same industry are collected, which overcomes the incommensurability of the financial indicators and the diversity of industrial standard in different industries. And then, using SPSS17.0and MatLab7.10., one completes the empirical analysis of the three models and obtains the fuzzy neural network with the smallest classificaion error. Specially, employing a kind of hybrid learning algorithm, one gets the globally optimum values of the parameters. This method not only reduces the search space but also improves the convergence speed. At last, one summrizes the advantange and shortage of the formulated models, which provides some methods and idea for future work.