Application Of BPNN And LR Combination Optimization Model To The Personal Credit Assessment Based On Feature Contribution

With the development of economic globalization and financial liberalization,credit assessment plays an important role in maintaining the normal relationship of social economy.It is necessary to establish statistic model for predicting personal credibility.Thus the customers can be discriminated as having good credit or having bad credit.However the mono-method-based models are not capable to simultaneously hold the robustness,interpretation and prediction accuracy of the models.Therefore,combination optimization of multiple methods has become an important developing for credit assessment.However,the variables of personal credit assessment are complex and diverse.In order to reduce the model complexity and improve the model prediction accuracy,it is necessary to consider reducing the dimension of the variables.In this paper,the methods of WOE-IV and principal component analysis were used to select variables.Then the personal credit assessment models were established by combining back-propagation neural network(BPNN)and logistic regression(LR).The records of customers' behaviors(January-June,2015)were collected from the transaction data of a communication corporation.The numerical variables were transformed using normalization and the classification variables were transformed to the binary classification form.The WOE-IV and principal component analysis were used to reduce the variables dimension,then modeling and forecasting with the selected variables.First,the LR models were established to calculate the probability of being bad credit for each customer,in order to discriminate that this customer has good credit or bad credit.The LR models based on WOE-IV prediction accuracy were 86.25% and 84.87% for the training samples and the test samples,respectively.The LR models based on principal component analysis prediction accuracy were 87.07% and 85.75% for the training samples and the test samples,respectively.Then,a BPNN model with one hidden layer was used to generate a new comprehensive variable for model optimization by tuning and selecting the number of hidden nodes.The optimal back-propagation neural network-logistic regression combination model(BPNN-LR)based on WOE-IV was established with 5 input nodes,7 hidden nodes and 1 output node.The model performance was slightly improved.The prediction accuracy was raised up to 86.33% and 87.96% for the training samples and the test samples,respectively.The optimal back-propagation neural network-logistic regression combination model(BPNN-LR)based on principal component analysis was established with 6 input nodes,8 hidden nodes and 1 output node.The model performance was slightly improved.The prediction accuracy was raised up to 86.16% and 88.32% for the training samples and the test samples,respectively.The results showed that the BPNN-LR model had higher classification accuracy than the LR model.It can provide technical reference to distinguish that customers have good credits or bad credits.And provide decision-making basis for corporation.