Application Of The Decision Tree Algorithm In Farmer’s Credit Rating Of Rural Credit Cooperatives

Rural credit cooperatives (rural cooperative banks, rural commercial banks, hereinafter referred to as "Cooperative") have long been used as the main force of rural finance, providing service to agricultural producers and small commodity producers. It plays an irreplaceable role in solving the farmers’ loans, increasing farmers’ revenue, supporting agricultural modernization and rural economic development. But its own NPL ratio is still5-7times the commercial banks. Therefore, objective, comprehensive and accurate assessment of rural customers’ repayment ability and willingness, refusing customers who don’t meet conditions will be a significant means of avoiding, controlling, and reducing losses. The traditional credit scoring systems are different in standards, with shortcomings of high costs, subjectivity, and low efficiency. Furthermore, nowadays with the social and economic development, extension and intension concept of farmers have changed a lot. Thus the traditional farmer credit evaluation systems are outdated. European and American countries’ experience shows that personal credit score can process customers’ loan applications fast, with characteristics of low cost, consistent standards. It is of great importance in the bank’s credit risk management. Modern credit scoring system in European countries widely uses statistics, operations research, artificial intelligence and other aspects of technology. Based on this, data mining technology has played a crucial role in the construction of credit scoring model.This paper firstly introduces different theories and methods of personal credit rating of both domestic and foreign scholars, and briefly introduces their advantages and disadvantages. Secondly, analyses are made to reveal the current conditions of credit rating for farmers. Qualitative evaluation in the rating process accounts for a large proportion, so establishing a quantitative one becomes important for reducing the risk of loans. Thirdly, this paper introduces the decision tree technology, the data mining method and the rating tool SAS. Fourthly, this paper sets up household credit rating model by collecting nearly four years of real peasant household loans sample data of the banking system, the data cleaning, conversion, sampling, analysis and using the decision tree algorithm in the SAS, and decision tree model is established, and the attribute assignment. Finally, the percentile system used by the credit scoring model in this paper is derived to prove that differentiated credit policy should be corresponding to different credit rating values.The innovation point of this paper lies in making the quantitative index of traditional credit rating from less than70%to94%, and improving the precision of the peasant household credit rating greatly. At the same time, this paper tests household credit rating model in high credit level of customers has higher prediction accuracy, while the precision of medium and low credit level of customers remains to be improved.