基于多维混合图和核心节点的社团发现算法

Community detection method based on multidimensional mixed graph and core nodes

社区发现和好友推荐算法一直是复杂网络的研究热点之一,在公安业务工作、网络舆情控制、电子商务等领域具有重要的意义。为解决公安工作中防范恐怖主义事件、打击易复发类犯罪、稳控重点群体等突出难题,论文提出了一种基于核心种子节点扩展的启发式社团发现算法。该算法通过有效融合多维信息形成混合图,以种子节点作为初始社区,综合考虑人物节点间不同交互行为和关联行为的权重,依托实战重点将协同过滤、Tanimoto系数、六度空间理论等算法相结合,最后把社区邻接节点中的活跃节点降序排列作为重点社团目标,得到了一种有核心节点的基于“人、事、地、网、组织”五维混合图的社团发现数据模型,为警务大数据及其他应用提供支撑。

Community detection and friend recommendation algorithms have always been one of the research hotspots in complex networks.It is of great significance in the field of public security business,network public opinion control,electronic commerce and other fields.To solve the problems of preventing counter-terrorism,combating the recurrence of crime and stabilizing key groups,this paper presents a heuristic community detection method based on core seed nodes extension.The method effectively fuses multidimensional mixed graph,and it considers user interaction and association behavior weight comprehensively.Seed nodes are used as the initial community.Relying on the actual combat,we focus on the combination of collaborative filtering,Tanimoto coefficient,Six Degrees of Separation and other algorithms.Finally,the active nodes in the neighboring nodes of the community are ranked in descending order as the key community targets.A data model of community detection based on five-dimensional mixed graph of people,events,places,networks and organizations with core nodes is obtained.It provides support for police big data and other applications.