基于支持向量机理论的中小企业信用风险预测研究

Research on the Prediction of the Small and Medium-Sized Enterprises’Credit Risk on the Basis of SVM Theory

商业银行实施巴塞尔《新资本协议》内部评级体系的重要任务之一就是估计债务人违约概率,而中小企业是公司风险暴露中的重要一类。因此,研究中小企业违约概率的估计并据此确定信用级别具有重要的现实意义。本文以200家中小企业作为样本,利用具有出色分类能力的支持向量机原理,找出了作为支持向量的财务指标,并引入新的违约概率计算方法,提出了企业个体与分类面的相对距离这一概念,利用该相对距离对企业的违约概率进行估计,进而通过违约概率来确定企业的信用级别。同时,对所给模型与现有模型进行了违约概率的一致性检验,得出了理想的结果。在此基础上,获得了相对距离与KMV模型中违约距离之间的关系。

During the process of carrying out the internal rating system referred in the New Basel Capital Accord, one of the most important tasks for the commercial banks is to estimate the default probability of the debtors＂ credit risk. The small and medium-sized enterprises are among the company risk exposure. Therefore, estimation of the small and medium-sized enterprises＂ default probability and further definition of the credit level based on this probability is of practical significance. This paper takes 200 small and medium-sized enterprises as samples and uses SVM theory which has extraordinary classify- ing ability to analyze a lot of financial indices of the samples. The paper reveals the financial indices that are defined as SVM, brings in the new method of computing the default probability, and puts forward the definition of the relative distance between the individual enterprise and the classification surface. It estimates the enterprises default probability based on this relative distance and defines the credit level according to this default probability. At the same time, this thesis carries out a consisten- cy test on the present model and the given model, with expected results achieved. In addition, it deduces the relations between the relative distance mentioned in this paper and the default distance in the KMV Model.