Design Of Early Warning Model Of Bond Default Risk Based On Random Forest

Affected by the central bank's "new rules for asset management" policies and the Sino-US trade war,the fierce competition in the bond market and the rise in corporate financing costs have led to an intensive outbreak of bond credit risk events.On the one hand,bonds are an important asset allocation for institutional investors such as commercial banks in China.The default of bonds cannot repay principal and interest,which will seriously damage the interests of investors.On the other hand,the problem of adverse selection leads to higher requirements for investors when investing in bonds.The rate of return is used as risk compensation,which in turn raises issues such as financing difficulties and expensive financing for SMEs.Therefore,the establishment of a bond default risk early-warning model to identify in advance the credit bond issuers that may be in default,is of great significance to institutional investors in optimizing investment decisions and the bond issuers to strengthen risk prevention.This paper uses a random forest model to model bond default risk,converts bond default risk into a more intuitive binary classification problem,and obtains a model with excellent performance through model setting and parameter optimization selection.For data imbalance,this paper uses the ADASYN method to optimize the data set.By comparing the model of the original data and the data optimized by ADASYN,the optimized data greatly improves the model's recognition of negative samples.Research shows that the random forest algorithm has high accuracy and recognition ability when making predictions,and is not prone to overfitting,so it is suitable for predicting and evaluating the risk of default.This paper designs a bond default risk early warning model based on random forest algorithm.The AUC index can reach 0.991,the accuracy of the total sample of the test set is 0.966,and the accuracy of the negative sample is 0.975.This paper applies a model of 8 bonds,which can make accurate judgments on the default conditions of bonds.Through the analysis of sample contribution,profitability mainly based on sales net interest rate and debt repayment ability are important indicators in the early warning model.Therefore,profitability and solvency based on specific indicators can be used as risk monitoring Key point.The research and application of bond default risk early-warning models should involve various public financial data and external environmental indicators.In order to make the model predictions more accurate and timely,this article recommends that supervisory authorities should increase the promotion of information disclosure mechanisms,core financial institutions should actively introduce advanced risk measurement technology,and other financial institutions strictly review credit issuance.Rating agencies need to maintain rating agencies.Neutral stance and rating level.Enterprises regularly use models to review their own risk situations.Investors need to use more scientific tools to identify bond default risks in time.