Research On Early Warning Of Credit Bond Default Risk Of State-owned Enterprises Based On Random Forest Model

Since its establishment,China’s bond market has remained relatively calm,and no default events occurred before 2014.However,this calmness is not due to the stability and benign development of China’s bond market,but rather to an unwritten rule in China’s bond market: rigid repayment.In particular,state-owned enterprises are often backed by local governments,providing implicit guarantees that make them highly trusted by investors.This situation has led to investors having great confidence in state-owned enterprise bonds,promoting the development of state-owned enterprise bonds.However,since 2015,state-owned enterprise bond defaults have occurred frequently,and in 2020,the "default wave" of state-owned enterprises emerged.Even AAA-rated state-owned enterprise "Huachen Group" defaulted on its bonds that year.State-owned enterprise defaults often involve huge amounts,which have a more severe impact on the bond market.Therefore,research on state-owned enterprise bond defaults is necessary.A review of related literature reveals several gaps in the field of bond default risk prediction.(1)there are relatively few studies that apply the random forest model to predict bond default risk,and these studies are mostly limited to financial indicators.(2)most existing studies focus on state-owned and private enterprises together,without specifically targeting state-owned enterprise bond defaults.(3)the samples selected in existing studies are mostly "ST" and "non-ST" companies,rather than actual default samples,which raises questions about the generalizability of the model based on this data.To address these gaps,this thesis proposes to establish a state-owned enterprise bond default risk warning model based on random forest.Actual default data will be collected,and an indicator system will be established from multiple perspectives,particularly by introducing a quantified indicator of government implicit guarantees.This will enable more specific analysis of state-owned enterprise bond defaults.This thesis examines state-owned enterprises that defaulted on bonds between 2015 and2021.To avoid subsequent defaults,the data for each enterprise is taken from the year prior to its initial default.Due to the small number of default samples,overfitting may occur,so the data is balanced using the SMOTENC method.The model parameters are optimized through learning curves and grid searching.Empirical results indicate that the random forest model accurately predicts state-owned enterprise bond defaults,with a 96.69% accuracy rate on the test set and an AUC value of 0.9678,demonstrating strong predictive and generalization abilities.The thesis also includes a case study of Huachen Group’s bond default,using various indicators from the year prior to the default to predict the outcome with the random forest model.The actual default matched the predicted results.Unlike the "black box" algorithms of traditional machine learning,random forest can output the importance of variables.By using the important variables output by random forest,we can evaluate the extent to which various indicators affect the results.In Huachen Group’s case study,this thesis first reviews the process of the company’s default,and then analyzes the reasons for its bond default from the perspective of the three important features output by random forest.The research shows that the more important variables come from three aspects:government implicit guarantees,enterprise financial factors,and external regulatory factors.At the level of government implicit guarantees,Liaoning Province has had slow economic development in recent years,and there is a phenomenon of economic water injection,which lacks a bottoming-out ability.At the level of enterprise self,Huachen Group is a "strong subsidiary,weak parent," with insufficient profitability of the parent company and unreasonable borrowing.These factors led to the company’s short-term repayment pressure being too high,which further resulted in the bond default.At the level of external regulation,credit ratings and audit opinions were both downgraded after Huachen Group defaulted on its bonds,and the lack of external regulation also contributed to the bond default.Regarding these three points,this article proposes three suggestions: Firstly,the government should transform its role,reduce implicit guarantees to state-owned enterprises,and shift from being a "savior" to a true regulator,fully exerting regulatory functions.At the same time,it should encourage the full use of market mechanisms,allowing enterprises to gain more opportunities and challenges in competition.Secondly,state-owned enterprises themselves should improve their own core competitiveness,focus on innovation,continuously optimize products and services to enhance profitability,and plan for debt issuance and repayment while enhancing internal control.Thirdly,intermediary agencies should strictly demand themselves,fulfill their duties well,break the "state-owned enterprise faith," and avoid the influence of implicit guarantees.They should strictly follow standardized workflows to execute various tasks and objectively provide audit or rating opinions.Through case analysis,this thesis aims to provide a reference for other enterprises and contribute to the creation of a good bond market environment.