The Development Of Credit Scoring And Its Application Based On Probit Regression

Credit risk as one of the most important financial risks, is the biggest issue of the current financial community. The credit industry in China is entering a period of rapid development. Many banks have been put forward to the grand blueprint of building retail banks, which makes the amount of credit and loans in all banks increasingly huge. The traditional manual credit has been unable to meet this demand, the credit scoring models emerged and developed in foreign banking and credit industry gradually will also be widely used in China. The research and model selection of the credit scoring become a challenging management issues in the bank community as the banking community is facing increasingly fierce competition. Similar to the logistic regression, the probit regression is a kind of generalized linear models, which can be used to solve Binary classification problem. In the establishment of credit scoring models, logistic regression is a very commonly used statistical method, while the probit regression is rarely addressed in this regard. In this paper, probit regression is used to establish apply for a credit scoring model to calculate the probability of default of each customer, and then the customers are divided into two categories, and the classification results of the model has been tested.In the first chapter, we described the necessity to develop credit scoring models in two aspects of the internal needs of the banks and regulatory requirements. Then we summarize the great advantages shown in the application of credit scoring models.The second chapter introduces the current development of credit scoring models in the domestic and foreign countries. Common methods of credit scoring models are described and discussed, and the advantages and disadvantages of these methods are compared. Finally we introduce the selection of the variable indices and commonly used method of data processing was introduced.The third chapter is the main focus. Firstly, we introduce the development process and the problems we face, which include the model classification, risk factor variables list, the definition and identification of "bad" samples, data sources.The fourth chapter is a complete modeling process. This chapter begins with extracting the modeling data from data sets in the form of1:1,2:1,3:1in accordance with good and bad samples. The original values and woe values of these data are used for modeling, while the remaining data samples are used for test. Finally, the results obtained in the various circumstances are compared with group and we obtain the best classification results.The fifth chapter is the concluding remarks, which summarizes the results obtained in the forth chapter, pointing out the deficiencies of the model and prospects for the future development of the probability of default model.