The Study Of Credit Risk Measurement Model For Small And Medium-sized Enterprises Based On SVM

Small and medium enterprises,which is the source of social innovation,play an essential role in the transformation an upgrade of China’s economic.However,SMEs has been plagued for a long time by difficulties of financing due to various reasons.The dilemma of SMEs’ financing difficulties is caused by many reasons,both for their own reasons,such as financial irregularities of SMEs,collateral difficulties,lack of credit and so on;and reasons about SMEs’ financing system,such as imperfect of risk assessment system of credit institutions,faultiness of the credit system used for SMEs and so on,of which the most fundamental reason is the existence of information asymmetry between banks and SMEs.An important reason for the asymmetry of information is that banks lack for credit risk measurement system designed for small and medium enterprises.As a result,establishing a measurement model to assess the credit risk of SMEs is an important way to solve the problem of SMEs’ financing.This article focuses on the subject of how to establish a suitable measurement model to assess the credit risk of SMEs.Firstly,it introduces the concepts of credit risk measurement and characteristics of SMEs credit risk and reviews the related methods,models of credit risk measurement.Considering that every method or model has advantages and disadvantages and the characteristics of SMEs,it noted that the appropriate method used for measurement of SMEs’ credit risk is credit scoring method.Method based on support vector machine,which is a machine learning method developing after statistical learning theory,is selected as a representative method of credit scoring.Support vector machine performs well when dealing with data with small sample and high-dimensional.Then,the matter how to use support vector machine to measure credit is elaborated.To measure credit risk,necessary process includes figuring out how SVM method works,selecting essential indicators for model,optimizing the parameters used in model and judging the effectiveness of model.Later on,data from real world is chosen to execute above-mentioned process.179 shares given special treatment are chosen as default sample.Accordingly,another 358 normal shares are chosen as normal sample.F-score assessment method and Wilcoxon rank sum test are used to filtrate essential indicators.Based on filtered indicators,parameters optimization process is executed,generating the optimal parameter combination.Finally,in order to evaluate the performance of model,performance of model based on SVM method and Logit model are compared.The result shows that,model based on SVM performs better than model based on Logit in terms of correct rate of the first type and correct rate of the second type.What’s more,model based on SVM has a better ROC curves.To build measurement model to assess credit risk of SMEs in china,machine learning technique,credit risk management theory and financial management theory are both used.As can be seen in the empirical results,the model performs well.