基于多元自适应回归样条的企业信用评估模型研究

发布时间：2018-05-07 12:49

本文选题：MARS + Logistic　；参考：《湖南大学》2012年硕士论文

【摘要】：现代市场经济是一种信用经济，企业作为市场经济中的主体，既是信用风险的承受者也是信用风险的产生者。随着我国市场经济的不断发展，企业的违约行为产生的越来越多，波及面越来越大，损害了投资者、债权人的利益。因此建立合理的信用评估体系，有利于企业掌握自身财务状况，便于内部管理与控制；有利于债权人事前预测债务人信用风险，保护债权人的利益；有助于引导社会资金向信用等级高的企业流动，促进了资金的优化配置；同时还促使监管部门更好地进行事前监控。从而有利于企业的健康发展，保护投资者、债权人的利益，提高市场效率，保障市场的有序运作。本文利用上市企业的财务数据，设计了信用分析的指标体系，利用多元自适应回归样条(MARS)方法对企业的信用状况建立信用评估模型。文章首先对现有的信用评估模型进行了分析、评价。其次，对MARS模型进行设计。再次，依据配对抽样以及上市企业2008年的财务数据建立了MARS模型，并与Logistic模型进行对比。最后，，基于研究结论提出了合理利用模型来防范风险的政策建议。通过理论分析及实证研究，本文得出了如下结论：首先，MARS模型作为一种非参数、非线性的回归方法，具有良好的拟合及预测能力，非常适合解决大数量的数据集问题，能够对非线性和变量间的交互作用进行建模，且运算十分便捷。其次，通过构建企业的MARS信用评估模型，并与Logistic模型进行对比分析，发现其拟合精度及预测能力均强于Logistic回归模型。再次，建立的MARS模型具有较广泛的应用价值。依据研究结论，本文提出要有健全的信息披露，以及关注企业财务数据的建议。
[Abstract]:Modern market economy is a kind of credit economy. As the main body of market economy, enterprise is not only the bearer of credit risk, but also the producer of credit risk. With the development of market economy in our country, more and more enterprises break the contract, which damages the interests of investors and creditors. Therefore, the establishment of a reasonable credit evaluation system is conducive to the enterprise to master its own financial situation, to facilitate internal management and control, to help creditors predict the debtor's credit risk in advance, and to protect the interests of creditors. It helps to guide social funds to flow to enterprises with high credit rating, and promotes the optimal allocation of funds. It also promotes better supervision and control by regulators. It is beneficial to the healthy development of enterprises, to protect the interests of investors and creditors, to improve the efficiency of the market and to ensure the orderly operation of the market. In this paper, the index system of credit analysis is designed by using the financial data of listed enterprises, and the credit evaluation model of enterprises is established by using multivariate adaptive regression spline mars (MARSs) method. Firstly, the paper analyzes and evaluates the existing credit evaluation model. Secondly, the MARS model is designed. Thirdly, MARS model is established based on paired sampling and financial data of listed companies in 2008, and compared with Logistic model. Finally, based on the conclusions of the study, the paper puts forward the policy recommendations of reasonable use of the model to prevent risks. Through theoretical analysis and empirical research, this paper draws the following conclusions: first of all, as a non-parametric, nonlinear regression method, Mars model has good ability of fitting and forecasting, so it is very suitable to solve the problem of large number of data sets. Nonlinear and variable interaction can be modeled, and the operation is very convenient. Secondly, the MARS credit evaluation model of enterprises is constructed and compared with the Logistic model. It is found that the fitting accuracy and prediction ability of the model are better than that of the Logistic regression model. Thirdly, the established MARS model has wide application value. According to the conclusion of the research, this paper puts forward the suggestion of sound information disclosure and concern about the financial data of enterprises.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.4;F279.23;F224

【参考文献】