Predicting Credit Spreads Of China's Private Placement Corporate Bonds Based On BP Neural Network

Non-public issuance of corporate bonds is also known as "private placement bonds".The issuers are mainly companies with low credit ratings,and issue bonds to raise funds within the scope of no more than 200 professional investors.The lead underwriter can only make private inquiries to potential specific investors when underwriting non-publicly issued corporate bonds.If the credit spread forecast for preissued bonds is used to determine an approximate interest rate range,the lead underwriter can use the interest rate range.Inquire with potential investors.Therefore,predicting a bond credit spread has certain reference basis for the sales inquiry of nonpublicly issued corporate bonds.Issuers and lead underwriters can determine bond pricing ranges based on predicted credit spreads and make sales recommendations with potential professional investors.Based on the existing research,this paper considers the company's asset size,financial leverage,corporate attributes,credit rating,issuance scale,issuance period,special debt provision,guarantee,lead underwriter reputation and accounting firm reputation as BP The input layer of the neural network uses the credit spread as the output layer to predict the credit spread of non-publicly issued corporate bonds.Empirical research finds that financial leverage,issuance scale,corporate attributes,credit ratings,special clause settings,guarantees,lead underwriters and accounting firm reputations have a negative correlation with bond credit spreads,and these factors will reduce bond credit spreads And the company size and bond maturity have a positive correlation with bond credit spreads,which will increase the bond credit spreads to a certain extent.Among them,the effect of company size,company attributes,credit ratings,guarantees and the reputation of accounting firms is more prominent.At the same time,this paper also uses a multiple linear regression model to predict the credit spread of non-publicly issued corporate bonds.It is found that the overall effect of the credit spread prediction of the BP neural network model on sample data is better than the multiple linear regression model.Finally,based on the conclusions of empirical research,we propose countermeasures and suggestions,hoping to effectively improve the bond market and help SMEs and non-state enterprises to reduce credit spreads.