A Research On The Farmer Credit Evaluation Based On BP Neural Network And Logistic Regression

The purpose of this paper is to analysis the various indicators, and foundsignificantly distinction between the higher credit rating farmers (good customer) andthe lower credit rating of the rural households (bad customer)indicators,byconstructing a model to calculate the farmers’s probability of returning the loan, thenobtain the credit scoring, finally get farmers credit rating criteria which for relevantparties to provide basis for decision-making and reference points.This paper presents the credit rating method commonly used in rural creditcooperatives of Hunan province firstly,then analysis the problems of current creditrating, on the basis of research achievements at home and abroad, this papercomparing the various methods of credit rating, and combining with the creditcharacteristics of farmers in china, choose BP neural network and Logistic regressionfinally, and buid hybrid model based on two kinds of methods considering theadvantages and disadvantages of the two rating methods, improving the predictionaccuracy and stability.Next, on basis of analyzing the farmer credit risk’s generation and its unique, wereach conclusion that our farmers credit characteristic is the combination ofindividuals and small and medium-sized enterprises, Therefore, in the construction ofindicators,we can refer to the indicators about individuals and small andmedium-sized enterprises’ credit rating at home and abroad. this article’s researchobject is the farmers who applys for loans in the part area of Hunan province, buildinginitial rating indicators separately from the following five aspects: the head of thehousehold and family members, assets, liabilities, business, family expenses.Weselected22indicators,in virtue of SPSS, we get12key indicators by using factorsanalysis and select646farmer families as sample, comparison with the farmer creditrating model based on the BP neural network and Logistic regression. Then establishcombination model by BP neural network and Logistic regression, the results showedthat: the overall accuracy is97.1percent, and the predictive rate of accuracy of goodcustomers is to be itself is98.8percent, the predictive rate of accuracy of bad customers is to be itself is81.8percent, its interpretability and soundness are ideal,so,this model achieves good prediction effect and has a certain application value.