一种新的基于粗糙集K.pdf

第 39 卷第 4 期 电 子 与 信 息 学 报 Vol.39No.4 2017 年 4 月 Journal of Electronics Information Technology Apr. 2017 一种新的基于粗糙集K-均值的社区发现方法 张云雷* 吴 斌 刘 宇 (北京邮电大学智能通信软件与多媒体北京市重点实验室 北京 100876) 摘 要：针对许多社区发现方法将社区看作一个集合而无法描述社区模糊区域的问题，该文提出一种基于粗糙集理 论的社区发现方法。该方法将社区看作两个集合，即社区的下近似集和上近似集，来刻画社区的模糊区域。该方法 首先选择 K 个节点作为社区的中心节点，然后根据节点与社区中心之间的距离将节点关联到社区中心节点形成社 区，接着重新计算社区的中心点及节点的社区标签，如此迭代直到收敛。通过公开数据集和仿真数据集验证了该方 法在社区发现方面的可行性和有效性。 关键词：社交网络分析；社区发现；粗糙集；K-均值 中图分类号：TP399 文献标识码：A 文章编号：1009-5896(2017)04-0770-08 DOI: 10.11999/JEIT160516 A Novel Community Detection Method Based on Rough Set K-Means ZHANG Yunlei WU Bin LIU Yu (Beijing Key Laboratory of Intelligent Telecommunications Software and Multimedia, Beijing University of Posts and Telecommunications, Beijing 100876, China) Abstract: Due to many community detection approaches regarding a community as one set of nodes which can not depict the vagueness of the community. A method based on rough set is proposed, it considers community as a lower and an upper approximation set which could depict the vagueness of the community. The method selects K nodes as the central nodes, then assembles iteratively nodes to their closest central nodes to form communities, and calculates subsequently a new central node in each community, around which to gather nodes again until convergence. Experimental results on public and synthetic networks verify the feasibility and effectiveness of the proposed method. Key words: Social network analysis; Community detection; Rough set; K-Means 1 引言 要分为基于图划分的方法、基于目标优化的方法、