The Construction And Empirical Research On Credit Evaluation Model Of Enterprises By Support Vector Machine

Small and medium-sized enterprises are playing more important roles in promoting economic growth,including job creation,cooperation with big companies and strong innovation abilities.However,most of the small and medium-sized enterprises face the problems of financial deficit,high financing cost and narrow financing channels.With the development of credit loans,needs of credit loan of small and medium-sized enterprises are increasing.The credit reference system in China is still imperfect.On one hand,financial management of small and mediumsized enterprises is not standard.On the other hand,commercial banks have not made adjustments on risk preference,management and the loan policy for small and medium-sized enterprises.Based on the situation,firstly,the article researches on the current bank credit business,including risk management,risk assessment,credit approaches and the shortcomings.Secondly,the article researches on the index system.It combines the existed financial indexes and newly raised non-financial indexes.Thirdly,the article calculates the weight of each financial index via AHP,by which it processes the data.Then it classifies the data by SVM.The innovation points lie in:Firstly,the research objects are small and mediumsized listed enterprises,which covers the shortage of previous studies;Secondly,it combines the existed financial indexes and two newly raised non-financial indexes;Thirdly,the article calculates the weight of each financial index via AHP,by which it processes the data.After analyzing,it finds that accuracy of credit risk forecast reaches 83.87% by using SVM(Only using financial data without using AHP);Accuracy of credit risk forecast reaches 85.48% by using SVM(Using both financial data and non-financial data without using AHP);Accuracy of credit risk forecast reaches 91.936% by using SVM(Using both financial data and non-financial data and using AHP).However,the accuracy of credit risk forecast only reaches 70.97% by using Logistic model withthe same data set.The results show that using AHP and SVM to process the data greatly improves the accuracy of personal credit evaluation.