Based On BP Neural Network Of Small Farmers Credit Risk System Evaluation

With the rapid development of industrialization of agriculture and the ruraleconomy, rural credit demand gradually increased. The main obstacle that hinderedthe financial institutions from entering the rural area is the credit status of thefarmers. So credit assessment of the farmers is a real way of bringing the troublesomeof farmers to get a loan to a change. Therefore, collecting the basic information of thefarmers’ micro-credit loans as a database, this paper tend to study the applicability ofthe model by constructing the credit risk assessment model and testing the creditstatus of the farmers.Firstly, starting from the study of theoretical basis of individual credit riskassessment and personal credit system, I made a basic theoretical analysis of the riskassessment of farmers’ micro-credit loans. Secondly, based on the detailed study ofartificial neural network model, this paper constructed a BP neural network model thatapplied to the farmers’ micro-credit loans and write a run programme of the BP neuralnetwork model by the Mathlab7.0software. Thirdly, seeing the data information ofthe micro-credit loans of farmers as a sample, I made80%of the data size as atraining sample to debug the risk assessment ability of the BP neural network,.Meanwhile, I use20%of the data size as a training sample to inspect the trainingresult of the BP neural network. After repeating sampling of the data for10times, Igot a ensemble of the risk assessment of farmers’ micro-credit loans of the BP neuralnetwork. The results of the accuracy rate of the BP neural network to estimate the database sample are: For the credit score, the training accuracy of the sample data is90.37%, and the test precision of the test sample is89.96%, the average error of thetest sample is0.03, the error standard deviation is5.07; For the credit rating, thetraining accuracy of the sample data is63.94%, and the test precision of the testsample is62.63%, the average error of the test sample is-0.01, the error standarddeviation is0.51.Finally, I made several policy proposals concerning the existing problems of thecredit risk assessment of the farmers’ micro-credit loans: Firstly, build an informationbase of farmers’ credit status through networking of the whole area; Secondly,strengthen the construction of the farmers’ credit assessment system; Thirdly, improvethe legal norm of farmers’ credit information; Fourthly, the government should play an active role in this area.