基于知识图谱的电信欺诈通联特征挖掘方法  

CALLING FEATURES MINING METHOD OF TELECOM FRAUD BASED ON KNOWLEDGE GRAPH

作　　者：凡友荣 杨涛[1] 孔华锋 彭如香[1] 姜国庆[1] Fan Yourong;Yang Tao;Kong Huafeng;Peng Ruxiang;Jiang Guoqing(Third Research Institute of Ministry of Public Security,Shanghai 201204,China;Wuhan Business University,Wuhan 430056,Hubei,China)

机构地区：[1]公安部第三研究所,上海201204 [2]武汉商学院,湖北武汉430056

出　　处：《计算机应用与软件》2019年第11期182-187,共6页Computer Applications and Software

基　　金：国家重点研发计划项目(2018YFC0830401,2018YFC0806903);公安部第三研究所2019年基本科研业务费专项资金项目(C19354)

摘　　要：提出一种基于知识图谱的通联特征挖掘方法,为电信欺诈案件相关的数据分析及线索挖掘提供技术支持.基于仿真的通话数据和电信欺诈案件数据,在分布式图数据库中构建知识图谱.在此基础上使用图遍历及图算法、混合高斯模型,从联系链路、必要人物、核心人物的发现以及社会关系识别这4个维度进行分析挖掘.在混合高斯模型中,提取9个关键通话特征,从通话模式聚类的角度来识别不同的社会关系.通过实验证明,图遍历及图算法能为电信欺诈人员和团伙的发现提供重要线索.混合高斯模型识别出了5类社会关系,并且发现涉案人员之间的通话模式具有一定的特殊性,即通话次数多且多发生在凌晨,通话时间较长且保持联系的时间较长.We propose a method of mining call features based on knowledge graph to provide technical support for data analysis and clue mining related to telecom fraud cases. Based on the simulation data of telephone calls and telecom fraud cases, knowledge map was constructed in the distributed graph database. Graph traversal and graph algorithm, and mixture Gauss model were used to analyze and mine the 4 dimensions of link, essential person, core person discovery and social relationship recognition. In the mixture Gauss model, 9 key call features were extracted to identify different social relationships from the perspective of call pattern clustering. Experiments show that graph traversal and graph algorithm can provide important clues for the discovery of telecom fraudsters and gangs. 5 kinds of social relations are identified by the mixture Gauss model. It is also found that there are some particularities in the mode of communication between the persons involved. The number of calls is more and mostly occurs in the early morning. The duration of calls is longer, and the time of keeping in touch is longer.

关 键 词：知识图谱 通联特征 线索挖掘 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]