The Research On Intelligent Early Warning For The Credit Risk Of The Commercial Bank Based On KPCA-PSO-SVM

Credit risk is one of the main risks that commercial banks face on.Once it happens,it will not only cause the loss of commercial banks,but also lead to the bankruptcy of commercial banks.Therefore,it has become one of the hot topics to analyze the credit risk of commercial banksand take effective measures to prevent and control credit risk in advancein the current theoretical and practical circles.Chinese commercial banks have a late start and short development time,so theyare lack of experience in credit risk management.At the same time,with the gradual opening of the Chinese capital market,foreign capitals come into Chinese market,which on the one hand speeds up the development of the capital market,on the other hand is likely to form a potential threat to Chinese fragile credit risk system of commercial bank.Therefore,it is of great significance for the development of Chinese commercial banks to optimize the early warning methods of credit risk,improve the management level of credit risk and consummate the credit risk management system.Based on the above analysis,this paper takes loan enterprises of the Chinese commercial bank,which is one part of listed companies in shanghai and shenzhen stock exchanges as the research object.Based on Chinese financial reality,the 16 variables that inducethe credit risk of commercial banks are selected and pretreated,Which produces 14 variables that can significantly distinguish credit risk and non-credit risk sample of commercial banks.And then the Kernel Principal Component Analysis(KPCA)method is used to extract the 14 variables with the aim of eliminating the high correlation between the variables.Support vector machine(SVM)artificial intelligence technology is introduced to construct SVM intelligent early warning model forthe credit risk of commercial bank.MeanwhileParticle Swarm Optimization(PSO)is used to optimize the parameters of SVM in order to carry out research work of credit risk early warning.The experiment results show that the KPCA-PSO-SVM model proposed in this paper has excellent prediction performance in credit risk prediction of commercial banks.The main contents of this paper are as follows:1.Study on the pretreatment of risk early warning samples.There are many problems in the construction of early warning model based on the sample's original variables,Sothis paper uses the normalization method and statistical analysis method to filter the original variables.The empirical results show that the normalization method can convert each variable as normal distribution,which can eliminate the dimensionproblem of variables.Meanwhile the statistical analysis method finds that the growth rate of business income and net profit growth rate can not significantly distinguish the credit risk and non-credit risk samples,which need to be removed from the variables.Through the experiment,this paper obtains variables,which are dimensionless and can distinguish the credit risk and the non-credit risk sample.2.Study on the extraction method of variables.There are a large number of indicators that can induce credit risk of commercial banks,which tend to be highly correlated.If we do not eliminate the high correlation characteristics of these variables and use them directly to model,it is easy to cause data redundancyand ultimately reduce the prediction effect of SVM intelligent early warning model.Therefore,this paper introduces the common principal component analysis(PCA)method and its improved method,the kernel principal component analysis(KPCA),to makes an empirical study.The empirical results show that the KPCA method in the variables extraction compared with PCA method is more efficient.At the same time,combined with SVM,KPCA can significantly improve the prediction effect of the SVM,but PCA method would decrease the prediction effect of SVM.It shows that the bank credit risk early warning variables have nonlinear characteristics,and the KPCA method is able to extract the nonlinear characteristics of the variables,which can effectively improve the early warning ability of SVM.3.study on optimization method of SVM parameters.The prediction ability of SVM intelligent early warning model depends largely on the penalty parameter and kernel function parameter.If the two kinds of parameters are not selected properly,it is very likely that the SVM model is over fitting or under fitting.Therefore,this paper studied the grid search method(GS)as the representative of the traditional parameter optimization method and genetic algorithm(GA),particle swarm optimization(PSO)as the representative of the heuristic algorithm in SVM parameters optimization effect.The experimental results show that the heuristic algorithm optimization method is better than the traditional one and PSO has more excellent prediction performance than GA,which can more effectively improve the prediction performance of SVM early warning model.Through a series of experiments,this paper believes that the credit risk early-warning modelof commercial bank based on KPCA-PSO-SVM is the best application tool for credit risk supervision departments of commercial banks.Regulators can use KPCA-PSO-SVM intelligent early warning model constructed in this paper to predict the credit risk of commercial bank in the future comprehensively and accurately and formulate and implement relevant policies and measures to immediately deal with credit risk in order to strengthen market supervision and effectively prevent credit risk.