基于集成学习的电信诈骗与营销呼叫行为识别方法研究  

Research on identifi cation methods of telecommunication fraud and marketing call behaviors based on ensemble learning

作　　者：郭向荣[1] 饶新益 胡琳欣[1] 纪纯妹[1] 薛飞[2] GUO Xiang-rong;RAO Xin-yi;HU Lin-xin;JI Chun-mei;XUE Fei(China Mobile Group Guangdong Co.,Ltd.Shantou Branch,Shantou 515000,China;China Mobile Group Guangdong Co.,Ltd.,Guangzhou 510000,China)

机构地区：[1]中国移动通信集团广东有限公司汕头分公司,汕头515000 [2]中国移动通信集团广东有限公司,广州510000

出　　处：《电信工程技术与标准化》2024年第10期30-34,45,共6页Telecom Engineering Technics and Standardization

摘　　要：传统的诈骗识别方法难以区分诈骗电话和营销电话,导致大量营销电话被误判为诈骗电话,降低了打击诈骗精准性。本文提出了一种基于集成学习的电信诈骗与营销呼叫行为识别方法。该方法通过对大量异常呼叫样本数据进行处理,提取出商业性营销呼叫和诈骗呼叫在不同场景下的呼叫特征参数,并利用集成学习技术对这些样本数据和特征参数进行训练和建模。实验结果表明,该方法能够快速、准确地判断出诈骗电话,从而实现对诈骗电话的限制和有效打击。Traditional fraud identifi cation methods struggle to diff erentiate between telecommunication fraud and marketing call,leading to a large number of marketing calls being mistakenly identifi ed as fraudulent,which reduces the precision of fraud combating.A method for identifying telecommunication fraud and marketing call behaviors based on ensemble learning is proposed.This approach processes a large volume of anomalous call sample data,extracting call feature parameters for both commercial marketing calls and fraudulent calls across various scenarios.Furthermore,it employs ensemble learning techniques to train and model these sample datasets and their characteristic parameters.Experimental outcomes demonstrate that this method can swiftly and accurately discern fraudulent calls,thereby facilitating the restriction and effective crackdown on such fraudulent activities.

关 键 词：电信诈骗 营销电话 集成学习 

分 类 号：TN918[电子电信—通信与信息系统]