基于核主元分析的带可变惩罚因子最小二乘模糊支持向量机模型及其在信用分类中的应用

信用分类是信用风险管理中一个重要环节,其主要目的是根据信用申请客户提供的资料从申请客户中区分出可信客户和违约客户,以便为信用决策者提供决策依据.为了正确区分不同的信用客户,特别是违约客户,结合核主元分析和支持向量机算法构造基于核主元分析的带可变惩罚因子最小二乘模糊支持向量机模型对信用数据进行了分类处理.在基于核主元分析的带可变惩罚因子最小二乘模糊支持向量机模型中,首先对样本数据进行预处理,然后利用核主元分析以非线性方式降低数据的维数,最后利用带可变惩罚因子最小二乘模糊支持向量机模型对降维后数据进行分类分析.为了验证,选择两个公开的信用数据集来进行实证分析.实证结果表明：基于核主元分析的带可变惩罚因子最小二乘模糊支持向量机模型取得了较好的分类结果,可为信用决策者提供重要的决策参考依据.

A KERNEL PRINCIPAL COMPONENT ANALYSIS BASED LEAST SQUARES FUZZY SUPPORT VECTOR MACHINE METHODOLOGY WITH VARIABLE PENALTY FACTORS FOR CREDIT CLASSIFICATION

Credit classification is one of the most important tasks in credit risk management and its main purpose is to provide credit decisions for credit granting institutions using the classification results. In order to correctly classify the different credit applicants, especially for default applicants, a kernel principal component analysis based least squares fuzzy support vector machine methodology with variable penalty factors integrating kernel principal component analysis and least squaresfuzzy support vector machine with variable penalty factors is proposed forcredit classification. In the proposed methodology, the sample data is firstpreprocessed, and then the data dimension is reduced by kernel principalcomponent analysis. Subsequently, the reduced data is used for classification analysis using least squares fuzzy support vector machine with variable penalty factors. For verification purpose, two publicly credit datasets are used for empirical analysis. Experimental results revealed that the proposed kernel principal component analysis based least squares fuzzy support vector machine methodology with variable penalty factors can obtain better classification results than other approaches listed in this study.