Study On The Credit Risk Prediction Of China Bond Market On CART

In recent years,China's bond market has developed rapidly.The financing method,financing channels and financing terms of the debt financing subjects have been constantly expanded and innovated,which makes the debt types of the bond market more diversified and the debt subjects more extensive.With the adjustment of macroeconomic situation and the growth rate of bond balance constantly hitting new highs,the risk of bond default is also accumulating.Since the rigid bond payment broke in 2014,bond defaults have been frequent,and the number of bond defaults hit a new high in 2018.Therefore,the effective prediction of bond credit risk has important theoretical and practical significance for the current financial risk prevention.With the data mining,artificial intelligence and other computer technology increasingly mature,and successfully applied to various fields.For investment institutions,it is necessary to apply machine learning to the credit risk prediction field of bond market through data mining technology to achieve efficient,sustainable,stable and healthy development.Based on the knowledge of financial management and data mining theory,we discusses how to use CART tree algorithm to build credit risk prediction model.In this paper,the structural characteristics of the factors obtained by CART model are used to better explain the risk characteristics of enterprises with bond market problems in recent years.At the same time,relatively reliable identification results are given.In the comparison of multi-model prediction,the overall accuracy of CART tree model is better,but the performance of recall rate relative to Logical Regression and Neural Network is slightly inadequate.However,in the feature analysis,CART tree algorithm can clearly find the enterprise problems,help investors quickly grasp the risk characteristics.Therefore,CART has a certain reference significance in practice.