摘要
银联商务作为国内最大的收单机构,服务于近800万的商户,针对目前赌博商户呈现高爆发、高增长态势,提出使用机器学习的方式通过商户的档案信息以及交易特征建立预测模型监测赌博商户。本文提出了组合Random Forest、GBDT、SVM和Logistic Regression的Stacking集成学习算法RGSL,实验结果显示,RGSL模型的预测准确率为97.57%,召回率为97.68%,相较于单一算法所构建的模型,效果有明显提高,达到了实际应用的要求。
China UMS,the largest acquiring institution in China,serves nearly 8 million merchants.In view of the current situation of high explosion and high growth of gambling merchants,this paper proposes to use machine learning to establish prediction model to monitor gambling merchants through merchant's file information and transaction characteristics.This paper proposes a Stacking ensemble learning algorithm RGSL,which combines random forest,gbdt,SVM and logistic regression.The experimental result shows that the precision of RGSL is 97.57%,and the recall is 97.68%,which is obviously improved and meets the requirements of practical application,compared with the model constructed by single algorithm.
作者
张野
叶国林
李欣刚
ZHANG Ye;YE Guo-lin;LI Xin-gang(China UnionPay Merchant Services Co.,Ltd.,Shanghai 201203)
出处
《数字技术与应用》
2020年第12期95-98,共4页
Digital Technology & Application
