等距特征映射的支持向量机模型在上市公司信用风险评估中的应用

The Application Analysis of SVM Model Based on Isomap in the Credit Risk Assessment of Listed Companies

为了实现上市公司信用风险的科学定量管理,提出了一种基于支持向量机(SVM)的信用风险评估方法。考虑到财务数据特征的非线性和高维性,采用等距特征映射(Isomap)算法对财务指标进行特征提取,以减少数据的冗余,针对人为选择SVM参数的盲目性,应用遗传算法优化其参数。通过以中国上市公司财务数据为基础的实证分析表明:基于Isomap的SVM模型比BPNN(BP神经网络)、PCA-SVM(主成分-支持向量机)模型具有更强的信用风险评估能力,小样本评估准确率达到91%。

In order to realize scientific quantitative management of credit risk in the listed companies, we put forward a kind of credit risk assessment based on support vector machine （SVM）. Considering the nonlinearity and high dimensionality of the financial data, we use isometric feature mapping （Isomap） to conduct feature extraction towards financial index to reduce data redundancy. In view of the selective blindness of SVM parameters, we use genetic algorithm （GA） to optimize the parameters. The empirical analysis on the basis of financial data of the listed companies in China shows that the SVM model of Isomap has a stronger credit risk assessment ability than the BP neural network and PCA-- SVM model. Its small sample evaluation accuracy reaches 91%.