基于隶属度的社会化网络重叠社区发现及动态集群演化分析  

Overlapping Community Detection and Dynamic Group Evolution Analysis Based on the Degree of Membership in Social Network

作　　者：国琳[1,2] 左万利[1,2] 彭涛[1,2] 

机构地区：[1]吉林大学计算科学与技术学院,吉林长春130012 [2]吉林大学符号计算与知识工程教育部重点实验室,吉林长春130012

出　　处：《电子学报》2016年第3期587-594,共8页Acta Electronica Sinica

基　　金：国家自然科学基金(No.60973040);国家自然科学青年基金(No.61300148);吉林省重点科技攻关项目基金(No.20130206051GX);吉林省科技计划青年科研基金(No.20130522112JH);中国博士后基金项目(No.2012M510879);吉林大学基本科研业务费科学前沿与交叉项目(No.201103129)

摘　　要：社会化网络中节点的复合属性可能为临时或过时状态,并且节点拥有一定能力维持固有状态,所以不可单纯依据新增数据或节点现有特征确定社区划分.本文提出可重叠社区发现算法及集群动态更新方案,根据网络历史数据分析节点对原始集群的隶属程度,并结合新增数据确定节点变化趋势,实现网络结构分析及社区动态更新.本文分别在不同数据集中测试聚类效果,实验结果证明算法既保持对新增数据的敏感度,也防止了节点短暂特征或节点维持固有状态的能力对划分结果的负面影响.The complex social attributes of nodes in the netw ork have a certain ability to maintain the former state,so it is inappropriate to determine community division merely based on new ly added data. This paper proposes an overlapping community detection algorithm and dynamic cluster update strategy,w hich,by fully analyzing historical netw ork data to compute the degree of nodes belonging to communities,determines the evolution tendency of nodes through incorporating incremental data to analyze the structure of the netw ork and update the division results automatically. Experiments on several typical datasets demonstrate that the algorithm not only ensures the sensitivity to incremental data,but also avoids the negative effect of temporary features in maintaining intrinsic states on the clustering results.

关 键 词：社区发现 社会化网络 聚类 重叠社区 自适应算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]