基于SenV-RBF的个人信用评分模型

将基于敏感性分析的RBF(radical basis function)网络应用于个人信用风险评估中,在训练中通过引入最大输出敏感度来度量隐藏神经元的数目及其径向基函数的中心,并构建了用于识别两类模式的基于SenV-RBF网络的个人信用评分模型.该模型对数据分布无任何要求,其在个人信用评分领域的运用,克服了统计等方法对假设较强的要求以及静态反映信用风险的缺点.经过比较分析,基于SenV-RBF网络的个人信用评分模型在分类的准确性和稳健性方面要优于传统的RBF,且精度可以达到支持向量机的水平.

Credit scoring model based on SenV-RBF

The radical basis function(RBF)networks based on sensitivity analysis was used in individual credit risk evaluation.In the training,the number of hidden neurons and the center of its RBF were measured by a maximum output sensitivity,consequently a credit scoring model based on the SenV-RBF networks was constructed to identify the two patterns.This model doesn't require any distribution of data and it overcomes,in the credit scoring area,the disadvantages of methods,such as too many requirements in the hypothesis and stationary reflection of credit risk statistical.Through comparative analysis,this credit scoring model based on the SenV-RBF networks is better than the traditional RBF in classification accuracy and robust,and its precision can reach the level to support vector machine(SVM).