The Application Of Data Mining In The Credit Risk Evluation

In the late1970’s, with the rapid development of China’s economy, our country consumer credit market has emerged, credit card consumption, personal housing loan, car loan and consumer durables loans, student loans and other forms of credit consumption gradually into people’s lives. The formation of consumer credit market is based on the premise of credit sales. In the process of credit sales, there is a certain degree of risk forever, namely customer no payment due or the customer can’t meet the payments due risk, we often refer to this risk for credit risk. So on customer credit analysis can help investors and Banks to choose reputable customers, eliminate risky clients, protect those customers having larger value customers for them, win the trust of the really high quality customers, as well as defend their own interests for banks and investors under the premise of minimizing risk.As for the study of credit risk evaluation model, researchers abroad mostly started from the lending market and the coiporate bond market, they usually select the bankrupt enterprise and the remaining enterprises, default loans with the default loans as sample to analyze the object, analyzed the index of financial trend characteristic and use it to establish credit risk assessment model, which is applied in credit risk evaluation. In China, its borrowing and bond market is still in the stage of development, the bankrupt enterprise financial information and lending market data is relatively difficult to obtain, there is no relevant information database as well, besides, and the performance and debt characteristics of an enterprise in the bank belongs to the bank internal secret, and it is very hard to get. By contrast, China’s stock market development is comparatively mature and developed. With the deepening of the economic system reform and the rapid development of national economy in our country, the expansion of listed companies’ numbers and size, has now become an important part of national economy. So most of the domestic research has done by using relatively mature stock market information, the financial situation of listed companies on China’s stock market information is important basis for evaluating enterprise credit risk.Based on the study of domestic and foreign references about credit risk measurement and data mining technology, this paper introduces the basic principle and the theory of credit risk of data mining technology, analyzes their advantages and shortcomings of several kinds of main data mining techniques. We put forward some methods on this basis to establishing the credit risk evaluation model of listed companies in China. Judging from the group of data with data mining method, this paper use the decision tree, Logistic regression method to establish credit risk assessment model, validation model. Based on the analysis of the important and financial index changes which can reflect the company’s main operation ability, predict the company’s credit risk estimate value, and for investors, Banks, or provide scientific decision basis for both parties.This paper is divided into the following several parts:The first chapter mainly introduced the selected topic background of this paper, analyzed the selected topic significance, then presented the literature review of the main achievements at home and abroad in the field of credit risk assessment, and lastly shows the idea and overall framework of this research.The second chapter mainly introduces the concept and characteristics of credit risk, and gives a one by one discussion and evaluation of main default probability estimation method which is included in modern credit risk assessment methods.The third chapter mainly introduces the data mining technology of decision tree, Logistic regression method, and their advantages and disadvantages, and puts forward the basic idea optimal weighted linear combination of modeling prediction.The fourth chapter is the empirical analysis, mainly introduced the research sample selection, index selection, data source, data processing and model establishment., model test and the result analysis.Chapter five is conclusions and outlook, obtained the main conclusions of this study, and summarized the lack of this study and future research directions and proposed the direction.