Research On Early Warning Of Credit Risk Of Small And Medium Enterprises Based On Decision Tree

In August 2018,the State Council held a meeting on the promotion of the development of small and medium-sized enterprises.It was pointed out that small and medium-sized enterprises contributed more than 70% of innovative technological achievements,provided about 50% of the country's tax revenue,and solved more than 80% of urban residents' employment.It is an important foundation for the high-quality growth of the national economy.In recent years,SMEs have encountered difficulties in financing and expensive financing,which have restricted the healthy and long-term development of SMEs.The reason why SMEs are caught in financing difficulties is because their high credit risk and high probability of default,which makes it difficult for SMEs to pass the loan review of commercial banks.Therefore,the article attempts to establish a credit risk early warning model for SMEs based on decision trees,which can more accurately predict the credit risk level of SMEs through the decision tree model,thus providing a reference for commercial banks to make loan decisions,reflecting the decision tree as a SME The credit risk early warning model has good application value.The article selects 44 companies with credit risk and 884 companies without credit risk on the SME board in 2018 as research samples,in-depth discussion on the status and characteristics of SME credit risk,revealing information asymmetry theory,credit rationing theory and complete contract theory have become the main cause of credit risk for SMEs.By comparing the credit risk early warning measurement methods,it shows that the decision tree model has the advantages of simple classification rules,relatively loose model assumptions and wider application scope.Therefore,the article chooses the decision tree method as a credit risk early warning model for small and medium-sized enterprises,and selects the most relevant 15 financial indicators from the six perspectives of profitability,operating capacity,cash flow,development capacity,relative value,and per share indicators through the filtering feature selection method.Because the sample size of small and medium-sized enterprises with credit risk selected in this article is relatively small,which belongs to the category imbalance problem,therefore,the RUSDT method is used in the establishment of the model,that is,the decision tree is used as the classifier,and the overall sample is imbalanced based on the random under sampling method.The data processing has overcome the problem of unpredictable samples that may cause a decrease in prediction accuracy.The empirical results show that the decision tree early warning model predicts the accuracy of SME credit risk levels as high as 95.5%.In order to further test the prediction performance of this model,this article compares the prediction effects of the Logistic regression model,SVM model and KNN algorithm with the decision tree model.Through the comparative analysis of the prediction results,we can see that the prediction accuracy of the decision tree model is the highest.In addition,in order to reduce the probability of SME credit risk as much as possible,it is necessary to focus on financial indicators such as profitability,operating capacity,and relative value to prevent financial troubles.