摘要
传统的信用评估模型,大多针对个人基本信息分析评估,忽略了交易维度的数据。针对这类情况,基于银行卡交易数据,提出一种基于交易数据的个人信用评分模型的方法。该方法主要包括三个部分:依据交易数据,提取有效的特征集;引入代价敏感随机向量,增强随机森林成员分类器的差异性;通过融合函数,输出信用评分。以互联网消费信贷场景作为研究对象,将该方法与经典算法进行比对。实验结果表明,该方法具有较好的评估结果。
The traditional credit evaluation model,mostly using personal basic information analysis and evaluation,ignores the transaction dimension of the data. In this paper,a method of personal credit score model based on transaction data was proposed by using the bank card transaction data. The method mainly included three parts： the effective feature set was extracted according to the transaction data; using Cost-sensitive random vector enhanced the diversity of random forest membership classifier; through the fusion function,model output credit score. This paper compared the results of the classical algorithm with the Internet consumer credit scene as the research object. The results of the experiment has good evaluation results.
作者
陈煜
周继恩
杜金泉
Chen Yu;Zhou Jien;Du Jinquan(China UnionPay Company Limited, Shanghai 200135, China)
出处
《计算机应用与软件》
北大核心
2018年第5期168-171,共4页
Computer Applications and Software
关键词
信用评估
交易数据
用户信用画像
随机森林
Credit evaluation
Transaction data
Credit personas
Random forest
