A Study On The Assessment Of Enterprises Credit Risk Based On Optimized Random Forest

With the gradual opening of Chinese financial market,prevention of financial risks is becoming an important challenge that each participant in financial markets must face.As one of the main financial risks,credit risk hinders the healthy development of financial markets.So,how to accurately predict the credit risk of enterprises has been becoming a hot issue with great practical significance.Based on the random forest model,this paper makes a research on the assessment of enterprise credit risk,and makes some optimization and improvement from two point of view.On the one hand,the SMOTE method is used to extend the original dataset,and the dimension of continuous variable is reduced by the method of variable grouping based on entropy criterion.On the other hand,the contribution function is introduced to decompose the original random forest prediction process,which quantifies the contribution of each variable in the prediction process.The study found that,after transformation of original dataset,the overall accuracy of the random forest model has increased.The results of ROC curve,CAP curve and K-S curve also show that the prediction performance of the model has been significantly improved,and the recognition speed of high-risk samples is improved.The results of variable contribution analysis show that the contribution degree of different variables of the same individual is different,and the contribution of the same variables in different individuals is also obvious,which can be help to identify key risk factors.The analysis of the variable contribution of the 7 default bonds shows that,the contribution of cost profit margin,return on total assets and sales net interest rate are larger,which significantly enhance the risk probability of the default enterprise.Base on the research conclusions,this paper proposes that,the intensity of domestic financial market information disclosure and market transparency should be improved,and financial databases should be enriched continuously,and the risk measurement technology should be optimized.