A Study On Credit Risk Measurement In China's Bond Market Based On Random Forest Algorithm

In recent years,China's bond market has developed rapidly.With the diversification of bond types and the expansion of financial entities in the bond market,bond financing tools,financing channels and financing conditions have also been expanded and updated.At present,China's credit bonds account for about 35%of the total size of the domestic bond market.Since 2014,when "11 Chaori Bond" defaulted as the first credit bond,the amount and number of defaulted bonds in China's bond market have increased proportionally.As of 2019,there are 145 defaulted bonds in China's bond market,involving 51 bond subjects.2020 In October,the successive defaults of state-owned enterprise credit bonds such as Yong Coal and Huarong brought market shocks.Although the number of bond defaults in China is low in terms of the total number of bonds,and the amount of defaults is only 1% of the total market value in terms of amount,however,under the current financial and bond market environment in China,the enterprises that issue credit bonds facing credit risks often include some systemically important enterprises,which will also bring about systemic financial risks once they default.Thus,in preventing systemic financial risks in China,forecasting the credit risk of bonds is an important part of the process.At present,there are still some problems in the methods and researches on the credit risk measurement of bond market in China as follows: firstly,at present,the relevant domestic institutions still mainly make subjective qualitative judgments in the process of bond credit risk rating;secondly,in other scholars' researches,there are fewer studies that include industry-related indicators and macro indicators for model construction,and in fact,these two often have a greater impact on domestic bond credit risk The second is that in other scholars' studies,there is less inclusion of industry-related indicators and macro indicators for model construction,which in fact tend to have a greater impact on domestic bond credit risk.To address the above problems,this paper conducts the following research: this paper adopts the random forest method,from more than 50,000 bonds in the domestic bond market from 2012-2019,the operation of merging and sampling bond subjects,and finally uses 1110 samples,in the selection of samples,for the reason that urban investment class enterprises have the hidden guarantee of government credit,this paper excludes the data of urban investment class companies.In the selection of the emergence of risk markers,this paper takes the decline of external rating of bonds as the sign of credit risk emergence.And the principal component analysis is used to downscale the index dimensions,while after extensive data preprocessing,the random forest model is constructed using Weka software to measure the credit risk of China's bond market and compared with other models.The results show that the overall accuracy of the random forest algorithm is good,but the model time consumption and recall rate are slightly lacking.At the same time,the model gives a series of quantifiable indicators to measure corporate credit risk,which is useful for investors to grasp the risk characteristics and has some reference significance in the specific practice of financial bonds.