Application Of Modified SVM On Small-and-Micro Enterprises Credit Risk Evaluation Of FX Bank

Credit risk evalution problem of small-and-micro enterprises(SMEs)is essentially a classification recognition problem with multi-indicator combined action.And the effective CRE contributes to increasing the obtained loans chances and easing financial pressure of SMEs,and can reduce loan risk and bad-debt probability of banks,so how to effectively evaluate the credit risk evalution of SMEs has important practical significance and application value both SMEs and banks.Thus,the credit risk evalution of SMEs in FX bank be seen as the study object,and further the techniques including the intelligent optimization algorithm,classification recognition of support vector machine(SVM)are used to systematically investigate the indicator system construction and the evalution classification model of credit risk evalution of SMEs.The penalty parameter and kenel parameter have significant influence and effect on the classification recognition performance and generalization performance of SVM,thus,for further improving predicting capability of SVM,a kind of recent harris hawks optimization(HHO)is imported into SVM and a modified SVM is proposed.Secondly,several techniques,which includs the consistency check of qualitative indicators,the correlation analysis and testing of quantitative indicators and the attribute reduction based on rough set,are applied to reduce and extract the key indicators of credit risk evalution of small-and-micro enterprises.Meanwhile,considering that key indicator data of credit risk evalution of SMEs has the small-sample,nonlinearity feature,a novle kind of credit risk evalution model of SMEs based on the modified SVM is construced.The study implies taht modified SVM based HHO has better classication performance and stronger robustness,the key indicators system of credit risk evalution of SMEs in FX bank is effectively constructed by the qualitative-quantitative selection and the attribute reduction based on rough set,and the credit risk evalution model of SMEs based modified SVM owns better identification accuracy of credit risk ranks and higere convergence efficiency.These above results have some expansion for credit risk evalution model of SMEs,and provide the corresponding technical support and theoretical support for the credit risk evalution of SMEs and credit-lending decision of FX bank.The thesis consists of 18 figures,10 tables and 54 references.