基于多分类器组合的个人信用评估模型

An Ensemble Credit Scoring Model Based on Multiple Classifiers

针对个人信用评估单一模型存在的不足,提出一种基于多分类器组合的个人信用评估模型。该模型综合了多元判别分析、logistic回归、神经网络、支持向量机等七种个人信用评估单一模型的预测结果,利用加权投票方法对其进行组合并输出最后预测结果。在某商业银行信用卡数据集上的测试结果表明,组合模型能有效地提高预测精度及稳健性,对信贷机构控制消费信贷风险具有很好的适用性。

This paper proposes an ensemble credit scoring model in which the results of seven base classifiers are combined through weighted voting,The accuracy and robustness of the ensemble model are tested using 5-fold cross validation with a credit card data sets from a commercial bank.Results demonstrate that the ensemble model based on multiple classifiers is efficient and predominant in credit scoring domain.