Research On Internal Rating System Model Of Credit Risks In Commercial Banks Based On IRB/SVM

In recent years, the banking industry has adopted more advanced risk control technique and approach and the risk management model has shifted from extensive management to intensive management, from depending on subjective judgment to depending on scientific quantitative analysis. After China's WTO entry, China's banking industry will have to open-up to the outside world and this is an inevitable trend. Chinese-funded banks are facing severe competition. We should be well aware that the focus of competition in banking industry has shifted from marketing to risk management. Good risk control ability and relevant potential for sustained development have become an important index of evaluating the banks. As an instrument for more effective control of credit risks, credit rating has been playing a very important role. But the current credit rating system in Chinese-funded banks is far from perfect that the banks are unable to get enough information about the borrower's credit. This makes it the top priority for us to establish an effective credit rating system which is suited to china's actual situation in order to reasonably and scientifically identify the banks'credit ratings and increase their predicting and controlling ability of business risks. This article uses an approach which involves both theoretical analysis and practice. It aims at establishing an effective credit rating model by using some advanced quantitative approaches.The content of the article is as follows:I. Making analysis on the current situation of credit risks system in commercial banks and the difficulties in applying IRB((Internal Ratings Based Approach) and enabling us to realize the drawbacks of our risk management system in commercial banks.II. Introducing the definition, main elements, structure of IRB so that we get to know more about the basic theories concerning IRB. Through the introduction to different credit risk measurement models.III.This article discusses the basic thinking concerning the construction of credit risk model in commercial banks and suggests the use of factor analysis technique during the model construction process. It also discusses SVM(Support Vector Machine)model based on IRB and creatively put forward an idea that we should support the use of SVM in judging whether there is financial fraud in the companies being rated and then use it as a risk weight to make the new rating system more objective and reasonable. This article tries to highlight some problems that we should think about while studying the construction of the risk measurement model, which is also the objective of the article. we have a further understanding of those models and the discussion on the merits and demerits of them also lead us to think about how to make good use of them.We have got the following research findings through the study in this article: We discuss the anti-fraud problem. Now we largely depend on subjective judgment to prove the validity of financial statements and it is difficult to distinguish those fraudulent elements involved as well as to set up an effective anti-fraud system. We could make analysis on the validity of the company's financial statements in previous years and use the result as a risk weighting function to calculate the PD (probability of default), thus making the system even sounder. An improved method of risk weighting is AHP ( Analysis Hierarchy Process), which is a quantitative analysis approach used more often than everThe research on the banks'credit rating system is a huge project. Though the author has done some research on it, but still found it hard to discuss it systematically and comprehensively. There are still some problems left unsolved. In terms of the construction of the model and domestic credit risk management system in the banking industry, a lot of data concerning the banks'customers'information and information about defaults are needed and this contradicts with the principle of confidentiality and security when it comes to internal banking data. Therefore it is unrealistic to construct an internal credit model with the effort of the author alone. What's more, because each bank has its own preference and objective when choosing the model, their focus might be slightly different. So what the author is doing here is to conduct some research on credit risk management system, focusing on the basic thinking and function realization of the model. That is to say, the construction of the model in next stage and the completion of the system should be a result of concerted efforts from the banks which supplies lots of data, manpower and material resources. In addition, to our regret, we are not able to incorporate code specific into the system due to the difficulties mentioned above.In the research area of the article, there are still a lot of questions left and many interesting subjects to be further discussed. The author will do more research on them in future work. Due to limited ability in study, there may be some improper ideas in this article, and the author welcomes helpful critical review from experts and scholars.