The Enterprise Credit Evaluation Model Research Based On Multivariate Adaptive Regression Splines

The modern market economy is a credit economy. As the most important subjectin a market economy, enterprises both bear the credit risk and generate credit risk.With the development of market economy in our country, more and more enterprisesdefaults have appeared and they have damaged the investors and the creditors’interests. Enterprise credit assessment can help the enterprises to understand theirfinancial status for internal management purposes; it can benefit creditors to predictwhether the debtor may be in financial trouble, and to estimate the debtor solvency; itcan guide social capital to flow to high credit rating business, thus to promote capitalallocation optimization; it can also prompte regulators to better monitor and control inadvance. In this case, it can be conducive to the healthy development of enterprises,protect the investors and the creditors, improve the market efficiency, and securymarket operation orderly.In this paper, the index system of credit analysis has been designed on thefinancial data of listed companies and the MARS method has been used to establishcorporate credit evaluation model. The article first evaluates the existing creditevaluation model. Secondly, the MARS model design. Again, a MARS model isestablished on the basis of the paired sampling as well as the financial data of listedcompanies in2008, compared with Logistic model. Finally, put forwardrecommendations of the rational use of the model to prevent risks based on theconclusions.Through theoretical analysis and empirical study, the article draws the followingconclusions: First, MARS model as a non-parametric, non-linear regression methodhas good fit and predictive ability, is very suitable for solving the problem of a largenumber of data sets, and is able to model the non-linearity and variable interaction,and the operation is very convenient. Secondly, by constructing the MARS creditassessment model and Logistic model for comparative analysis, it has been found itsfitting accuracy and predictive ability is stronger than the logistic regression model.Again, the establishment of the MARS model has a wider application value. Based onthe findings, the paper proposes to have a sound information disclosure, and payattention to the corporate financial data.