Personal Credit Assessment Based On Support Vector Machine

Personal credit assessment is an important part of the credit business.The bank assesses credit default risk through analyzing the loan application submitted by the applicant,and thus decides whether to authorize personal loan applications.At the same time,the accuracy of the risk discrimination model directly affects the bank's credit risk.Credit assessment is to evaluate the borrower's credit quality based on some customer classification methodologies.Actually,the classification algorisms have been studied for over 60 years,from statistical modelling to non-parametric methods and further to artificial intelligence approaches.Scholars are committed to create more accurate and efficient assessment method.Support Vector Machine(SVM)is a kind of new machine learning technology based on Statistical Learning Theory.It is widely used in dealing with small sample,nonlinear data and high dimensional pattern recognition,showing the specific advantages.This paper aims to construct the SVM personal credit assessment model,and to discuss the feature selection and parameter optimization in SVM model establishment process in order to obtain higher prediction accuracy.Particularly,this paper reviews the principle and algorithm of SVM classification model and then finds that the traditional SVM model cannot filter the predictive variables.Thus,we put forward judging the importance of credit index by calculating its Information Value,and the selected variables are put into the SVM model based on this method.At the same time,related research shows kernel function parameters and penalty factors in SVM model plays a significant role in the performance of the model.In this paper,the Swarm Intelligence optimization algorithm is combined with support vector machine(SVM),respectively using three Swarm Intelligence algorithm of genetic algorithm(GA),particle swarm optimization(PSO),Grey Wolf Optimizer(GWO)to optimize the SVM personal credit assessment model,and accordingly construct combination optimization personal credit assessment models of GA-SVM,PSO-SVM and GWO-SVM.We test the model by investigating the real credit data,and and using the Matlab programming.Our tests focus on comparing the prediction accuracy of the combined SVM model with selected variables and the three combined optimization models.It is proved that both the deletion of the redundant variables and the key parameters optimization of the SVM model can significantly improve the classification accuracy of the optimization of the model.The method proposed in this paper has a good implication in dealing with credit quality assessment and can provide effective decision-making advice for credit departments.