摘要
电信客户信用风险等级评估是对电信客户的信用风险进行等级分类.针对建立客户信用风险等级分类模型时,大量带有类标注数据难以获得的问题,提出了基于主动学习的分类器建模方法,并对基于QBC(委员会投票选择)的主动学习算法进行改进以提高分类器的预测精度.通过对实际电信客户数据进行信用风险等级建模实验,结果表明:应用新算法,分类器使用了较少的带类标签样本数据,达到了与被动学习相同的精度,大大降低了信用专家评估数据的工作量.
Evaluating telecom clients＇ credit risk rate is classifying their credit risk level. An approach based on active learning was proposed for solving the insufficient labeled data problem in building a credit risk rate classifier. The new QBC （query-by-committee, QBC） method of active learning was presented to improve the classifier＇s accuracy. By applying the actual telecom clients data in the experiment, the results show that the model built by the new algorithm with less labeled training data can reach the same accuracy as passive learning. This can reduce annotation cost for credit evaluation experts.
出处
《北京科技大学学报》
EI
CAS
CSCD
北大核心
2007年第4期442-446,共5页
Journal of University of Science and Technology Beijing
基金
北京市教委重点学科共建项目资助
国家民委"十一五"科研项目(No.07ZY07)
关键词
电信客户
信用等级
主动学习
投票
相对熵
telecom clients
credit rating
active learning
vote
Kullback-Leibler divergence
