基于后验概率的个人信用评估SVM模型

Personal Credit Scoring Based on Posterior Probability SVM

针对标准SVM模型在个人信用评估中单纯将消费信贷申请者划分为违约类或未违约类的不足,提出了利用基于后验概率的SVM模型进行个人信用评估的方法。利用商业银行的消费信贷数据进行的实证研究表明,基于后验概率的SVM模型通过将标准SVM的决策值转化为后验概率输出,能够对样本属于未违约的概率进行估计,并能够据此划分信贷申请者的信用等级,对于商业银行根据不同的经营目标制定相应的信贷政策更具有实践意义。

Aiming at the insufficiency of standard SVM used for personal credit scoring merely to classify the con- sumer credit applicants into default or non-default group,this paper presented a method for personal credit scoring by using posterior probability SVM model.Using consumer credit data to cheek the model s applicability,the re- suits indicate that posterior probability SVM can output the posterior probability by transforming the decision value of standard SVM model.The posterior probability can be used to...