Commercial Bank Personal Credit Evaluation Models Study

Personal credit evaluation is a key link to commercial banks personal financial business development and credit approval, is the core of the personal credit risk management. At present, China’s commercial banks to conduct personal credit evaluation, in most cases are still based on the credit staff subjective judgments and qualitative analysises on the customer, there are low efficiency, high cost and low accuracy shortcomings, already can not meet the need of rapid development of personal financial services of commercial banks. Introducing the quantified, automated personal credit evaluation system, is a urgent need of the commercial banks for enriching business, developing market and improving the competition ability.In this paper, on the basis of summarizing and drawing on the relevant research results and practical applications, the paper use three methods to build respectively three single models for personal credit evaluation:Logistic regression model, Analytic Hierarchy Process Model and RBF neural network model, and then use credit data to carry on a empirical analysis and comparative study on them. By combining with the actual situation in our country, this paper further built a combined model which based on the analytic hierarchy process method and the logistic regression model. This paper argues that the Analytic hierarchy process-Logistic regression combined model not only to retain the analytic hierarchy process model’s advantages of easy modeling and strong operability, but also absorb the Logistic regression model’s advantages of rigorous calculation, higher forecast accuracy, better stability and wide applicability. This modeling approach can maintain a relatively high prediction accuracy while reduce the workload of data collection and simplify business process. This paper believes that the combined model is applicable to the actual situation of our contry in which personal credit registry system is not perfected and personal basic information database is incomplete.