The Personal Credit Evaluation Model Based On Data Mining Technology Research

There has been rapid development of the personal consumption loans in China recently. With the development of it, however, the credit risk problem exposed on the business has threaten the development of consume credit market, and the commercial banks and its regulators have to pay more attention to the personal credit risks. Nowadays, personal credit risk evaluation which basically decided by experience or qualitative analysis reveals more disadvantages compared with commercial credit rating. Facing the surge of loan portfolios and the shortage of employees, it may result in prolonging approval, increasing of faults, declining of service standard and risk control ability, thus leads to deterioration of asset quality, the loss of potential clients, and the decline of competitiveness. This dissertation develops a personal credit evaluating model for the purpose of improving efficiency and accuracy of lending, and enhancing the asset quality and risk control ability. In this case, the comprehensive, systemic research on Individual consumer credit theory, consumer credit risk theory and individual credit evolution is of important theoretical value and practical significance.Although many positive attempts are done, the development and application of personal credit assessment model in Chinese bank industry is still in its infancy. The research of accuracy and applicability of different models also requires in-depth study. At present, Chinese personal credit rating system is imperfect and the personal credit information of commercial banks is still incomplete. Under such circumstances, it is very meaningful to develop a model that has reference value to the practice.In the dissertation, different methods applied to personal credit assessment are studied. The methods are comparative studied by use of experiential data. The experimental analysis shows that the Logistic regression model is the most suitable model which is more stable than other models, and its forecasting accuracy is higher.This dissertation is divided into five chapters:Chapter One introduces data mining theory and its applications in credit assessment. Firstly, it explains the basic concept of data mining. Then it introduces the function of data mining as well as the methods. Subsequently it discusses the basic procedure of data mining. Finally, it discusses data mining applications. Data mining technology plays an important role in commercial bank risk management.Chapter Two discusses the basic concept of personal credit assessment as well as it development history. Personal credit assessment is to judge the creditworthiness of a credit applicant. In western countries, the method of quantitative analysis is widely used. The author summarizes several popular methods and the main achievements of predecessors in recent years. It is necessary to make a deeper research in developing more accurate and more suitable models for our practice.Chapter Three discusses the problem of preparing data before developing models. The data used in the dissertation is real data from personal client history information database which is provided by certain commercial bank in Sichuan province. The practical data include incomplete, inconsistent, imprecise or repeated data, thus data clean and data transform are necessary. Then, in the end of this chapter, the author introduces the sampling method used in the dissertation.With the aid of real personal client’s data and SAS software, Chapter Four develops personal credit assessment models by using the following methods:Discriminate analysis, Logistic regression, K-Nearest neighbor, Decision tree. The author introduces the procedures in details and calculates the total misclassification rate and the two kinds of misclassification rate of training sample data and validation sample data for each model.Chapter Five concentrates on the assessment and comparison of the several models developed in the previous chapter. The result is that the Logistic regression model is the best model which is worthy promoting in the practice. Moreover, in the end of this chapter, the author puts forward several issues which should be pay attention to during the application.The dissertation has some new ideas:(1)The dissertation uses the real personal credit data to empirical analysis. It discusses the pretreatment methods and procedure of real data.(2)The dissertation uses several methods to sample data in order to find the relationship between clients’credit level and clients’identity information. Then comparative study of methods is used to assess the models. The final model recommended is of high objectivity, and high accuracy of forecasting and so on.(3)The dissertation uses many non-parametric analysis methods. These methods have no special requirements for data distribution, so they can avoid the setting difficulty of traditional model. The thesis discusses these methods and provides alternative methods and referential means which may be helpful for personal credit risk management.