计算机应用研究2018，Vol.35Issue(2)：421-427,7.DOI:10.3969/j.issn.1001-3695.2018.02.022
基于Ext-GBDT集成的类别不平衡信用评分模型
Class-imbalance credit scoring using Ext-GBDT ensemble
陈启伟 1王伟 2马迪 3毛伟1
作者信息
- 1. 中国科学院计算机网络信息中心,北京100190
- 2. 中国科学院大学,北京100190
- 3. 北龙中网(北京)科技有限责任公司,北京100190
- 折叠
摘要
Abstract
In view of class-imbalance and cost-sensitive problem in real credit scoring business,as well as financial institutions prefer to assess credit risk of the loan applicant in an intuitive way,this paper proposed an Ext-GBDT ensemble model for class-imbalance credit score.In this proposed model,firstly,it adopted an under-sampling method to randomly samples several subsets from credible customer(the majority class) and then combined each of them with default one (the minority class) for generating several class-balance training subsets.Secondly,it emploied different training subsets as well as feature sampling and parameter disturbance method to train several diverse Ext-GBDT models.After that,it integrated the predicted result of different models by using the simple average method.Finally,it transformed credit probability into credit scoring.In terms of AUC and cost-sensitive error rate,this model againsted five well-known credit scoring models and their ensemble model on UCI German credit dataset and the research results reveal the validity of the proposed method.关键词
信用评分/类别不平衡/代价敏感/Ext-GBDT/集成学习Key words
credit scoring/class-imbalance/cost-sensitive/Ext-GBDT/ensemble learning分类
信息技术与安全科学引用本文复制引用
陈启伟,王伟,马迪,毛伟..基于Ext-GBDT集成的类别不平衡信用评分模型[J].计算机应用研究,2018,35(2):421-427,7.
