Research On Credit Risk Assessment Of SMEs Based On Lasso-Cox

Affected by the outbreak of covid-19 epidemic,our country adopted strict control to deal with the diffusion of the virus at the beginning of the year,which led to a large number of enterprises' bankruptcy because of the inactive production.Because of its defects on its operation and the influence of the epidemic,the survival road of SMEs is being more and more difficult.The high cost of financing and the close restraint of financing are the main factors,which limit the rapid expansion of such SMEs.Therefore,we should establish a scientific credit risk assessment system,and commercial banks can use this system to make better assessments of the credit risk of these companies and reasonable pricing of financing costs,which is an effective way to relief the financing stress of SMEs.Based on the background of the financing difficulties which SMEs are faced,this paper makes theoretical research from five aspects,such as credit risk theory,credit risk theoretical models,influence factors,evaluation index selection and evaluation model selection.And according to the characteristics of SMEs and the research results of scholars,this paper uses the internal asset financial ability index and sustainable development ability index to construct the static credit risk evaluation index system of SMEs,and select a total of 35 corresponding indicators.Because of the unbalanced distribution of the sample data of SMEs,the proportion of default enterprises is too small,and the imbalanced proportion of the positive and negative sample will reduce the credibility of the prediction model to a certain extent,so SMOTE model can be used to expand the default sample,and then ensure the balance of the sample.In addition,according to scholars' research,when there are multiple indicators,there is often a problem of redundant variables.The existence of redundant variables will affect the operation of the model,but the Lasso model can handle such variables better,so this paper considers the establishment of Lasso-Cox model to find the effective indicators of credit risk assessment of SMEs and carry out credit risk assessment.In order to illustrate the effectiveness of the Lasso model,this paper further uses the PCA model and random forest to reduce the dimension of credit risk evaluation index system,and combines the evaluation model for accuracy comparison,from different dimensions to verify the effectiveness of the Lasso-Cox model.Finally,the Cox model is very particular and macro-economy will have a great impact on the survival status of SMEs,so this research adds three macro variables for dynamic research and analysis.And the empirical results show that:(1)the Cox model can indeed produce more accurate evaluation results than the traditional model,(2)Lasso performs better in identifying valid variables,and adding macro-factors can increase the predictive accuracy of the Cox model,and(3)innovation factors,the quality of enterprise personnel and basic characteristics of enterprises are also important in the credit risk assessment of SMEs.And the results of the study also show that the Cox model is suitable not only for the medical field,but also for credit risk assessment for SMEs.Finally,the wide application of this kind of assessment method can help to alleviate the lack of credit system and the difficulty of borrowing caused by the defects of SMEs,and promote the steady development of our country's social economy.