Application Of Integrated Support Vector Machine On Private Enterprise Credit Bonds

As an important part of the socialist economy with Chinese characteristics,private enterprises have made indelible contributions to Chinese economic development.However,compared with state-owned enterprises which often have government support,private enterprises are often discriminated by banks because of their own credit qualifications.Since the fourth quarter of 2016,the macroeconomic market has experienced three stages including "financial de-leverage","financial stable leverage with economic de-leverage" and "economic de-leverage".The strict supervision of the government have a considerable impact on the financing and operation of various enterprises.Some private enterprises previously expanded their debts too much through the issuance of bonds,hoping that they can repay their debts through their own operating profits or refinancing.However,with the gradual expiration of bonds markets,the corporate capital turnover have led to the inability to repay previous bonds,which ultimately led to a large-scale "default wave" in the bond market.In the field of risk assessment,traditional quantitative analysis relies mainly on individual subjective judgments which often lead to inaccurate results.However,with the development of relevant model theory,the relevant characters are gradually quantitatively analyzed,and some models are proposed such as linear discriminant,logistic regression model and neural network model.In the 1990s,based on statistical learning theory,Vapnic proposed a support vector machine model.This model not only has small sample learning ability,but can also solve traditional problems such as nonlinearity,high dimension and over-fitting problems effectively.Ensemble learning generates a number of differentiated sub-learners through certain algorithm training.and makes accurate decisions which can improve the generalization of the model.This paper selects the private enterprises which have publicly issued credit bonds in the market as the research object,using the company's financial data to predict its future bond defaults.At first,we select 16 characteristic variables from different dimensions of the company's financial indicators,and then reduce its number by logistic regression.For the models,we firstly use the support vector machine model to determine its optimal kernel function.Then we introduce two common integrated learning methods such as:Bagging algorithm?Boosting algorithm and then respectively perform classification prediction.Finally,we improve the support vector machine model through the introduction of Bagging algorithm and prove that the model has more accurate prediction.