The Empirical Research On Bonds Using Lasso And SVR

With the rapid economic development,China's bond market has continued to expand.Bond issuance has increased sharply,total amount of bond issuance has continued to increase,and the bond market has developed rapidly.At the same time,due to the impact of factors such as reduced product demand and macroeconomic adjustments,some companies cannot fulfill debt obligations as promised,thereby forming a default.This makes the listed company's default warning and credit rating effective references for investors' investment decisions.This article will study the default of listed companies' bonds based on the two methods of Lasso and SVR.On the one hand,by comparing the advantages and disadvantages of the Lasso method and the Pliable Lasso method in the selection of sample data variables,the Lasso method is used to select variables for the financial data of listed companies.By using the Aalen additive risk model constructed by means of the survival analysis method to propose Probabilistic Lasso-Aalen model to study the relationship between the default probability of listed companies and various financial indicators.Research shows that several indicators such as total assets,retained earnings rate,net asset return,asset-liability ratio,cash income ratio,etc.all affect the strength of the company's default,that is,these indicators are robust;the total cash debt ratio,the total assets,and the return on net assets are all highly time-varying;and the greater the value of the retained earnings rate and the cash income ratio,the lower the probability of the company defaults.The research method in this paper provides an important theoretical basis for the default prediction of listed companies.On the other hand,the exploratory factor analysis method is used to determine the model as a non-linear support vector,and the support vector machine and support vector regression are used to classify and predict the credit bond rating.At the same time,by comparing the different kernel functions of the support vector machine,we can select the the most appropriate kernel function.The study found that in the sample data,AA+ grade bonds have the best prediction effect among all bonds;in the issue of bond credit rating,it is better to use radial basis kernel function for prediction.This article uses the Lasso-Aalen model and the support vector regression model to study the possibility of default of listed companies and the credit ratings of corporate bonds and corporate bonds of the Shanghai and Shenzhen exchange markets,which further illustrates the feasibility and advantages of machine learning methods in the empirical research of the bond market.