基于BP神经网络的农户小额信贷信用风险评估研究

利用2009年杨凌区三家农村信用社的实地调研资料进行了农户小额信贷信用风险评估的实证研究,对指标变量分别进行正态性检验、差异性检验和多重共线性检验,利用MATLAB7.0软件建立了8—14—1结构的BP神经网络农户信用风险评估模型。模型对训练集样本的总体判别正确率为100%,对测试集样本违约类农户的预测正确率达90%,总体正确率达84.09%。准确度较高,能够为农村信用社识别农户信用风险提供较好的依据。

Research on Assessment of Credit Risk of Small-amount financing for Farmer Households Based on BP Neural Network

Based on the survey data of three rural credit cooperatives’ in Yangling of Shaanxi in 2009,this paper sets up a BP neural network model of credit risk assessment which has a 8-14-1 structure with MATLAB 7.0 software to study small-amount financing for farmer households,and every variable is identified by normality test,variance test and multi-collinearity test.The accuracy rate of this model is about 100%for training set,90% for testing set,and 84.09%for the total.So this model can provide a good basis for RCC to identify farmers’ credit risks.