【摘要】 针对个人信用评估中数据海量性以及与影响因素之间的非线性问题,利用最小二乘支持向量机(LS-SVM)中基于GCV准则和Newton-Raphson算法的正则化参数快速选择方法建立新的个人信用风险预测模型.并把该模型与Fisher线性判别分析、Logistic回归以及半参数广义可加模型的判别效果进行了实证比较分析.结果表明该方法不仅具有快速高效的模型选择能力,并且具有较优的判别预测能力.更多还原

【Abstract】 In view of massive individual credit data as well as non-linear relation between credit and its influencing factors,a new individual credit risk forecast model is established using LS-SVM,in which regularization parameter is selected based on the GCV criterion and the Newton-Raphson algorithm.Distinction effects are compared among the new model,Fisher linear discriminant analysis,Logistic regression as well as generalized additive model.The results show that the proposed method not only has the effective model selection ability,but also has the superior distinction predictive ability.更多还原