Credit Scoring System Based On Data Mining Analysis And Design

Currently as the globalization of economy and information is accelerating,the computerization of information of the banks in China has made substantial achievements. Some operational information systems have been constructed. However,since large amounts of business data are saved in the databases of banks,it is difficult to analyze these data while making business decisions. With the aid of the data mining technology,we can find the rules and modes that underlie the existing business data of banks in order to set up some decision-making models to develop DSS of banks. This is one of the focuses of electronic information construction of the banks in future. At present China lacks a relatively matured personal credit system, so it is difficult for banks to evaluate the credit states of the personal clients accurately. In order to minimize the risk,banks emphasize the second fund source offered by debtor,such as mortgage and so on. This situation not only deviated from the ultimate purpose of the personal loan service,but also restricted the further development of the service. The objective assessment of the credit states of personal clients has become a key issue of banks in China to launch personal loans. In western countries,banks generally adopt the method of personal credit scoring to appraise the credit state of the personal clients quantitatively. Compared with the method of subjective judgement,the method of personal credit rating is more comprehensive,more scientific, more objective and better justified. Western countries mainly adopt the technology of the data mining to develop personal credit analysis models to help in the personal loan decisions of the banks. Generally speaking, personal credit analysis models include personal credit scorecard model and personal client categorization model. When the characteristic value ofa personal client is set,the first model can give a credit score to appraise his credit state. The second model can classify the personal clients into categories (such as the category of good clients or the category of bad clients) on the basis of the personal credit scorecard model. After developing preliminary personal credit analysis models,some test data must be chosen to appraise the effectiveness of models,and repeated adjustments are needed to improve them. The personal credit scoring system based on personal credit analysis models can offer scientific decision-making methods to the banks and improve the efficiency and profits of banks. This thesis first explains the basic theories of the personal credit scoring and the main techniques of data mining,and then analyzes the main procedure of developing the personal credit scoring system by the data mining. It summarizes the main achievements that western countries have made in developing personal credit scorecard model and personal client categorization model. The author also adopts the data mining tools of SAS to analyze the personal clients' information offered by Sichuan Province branch of the Agricultural Bank of China to develop personal credit analysis models and to realize the analysis and designing of personal credit scoring system. This thesis is divided into four chapters. Chapter One focuses on analyzing the basic theories of personal credit scoring and the present situation of the personal credit scoring. This chapter first explains the basic concepts of the methods of both personal credit scoring and subjective judgment to demonstrate the main advantages of the former. Then it discusses the development of the personal credit scoring in western countries. Then it analyzes the current situation of personal credit scoring in western countries,pointing out that analysts mainly adopt the technology of the data mining to develop the personal credit scoring system,and introduces the commonly adopted personal credit scoringmethod—FICO. Finally,it analyzes the current situation of personal credit scoring in China,and again emphasizes on the main advantages of the technology...