摘要
社区挖掘是复杂网络分析中一项重要工作.目前已有许多好的社区挖掘算法,但这些算法大多基于节点间的连接关系发现内聚的社会团体,而实际网络中节点大多具有不同的行为和影响力.基于此,充分考虑社区内节点相互连接紧密以及节点具有不同影响力的特性,提出一种基于极大完全图扩展的社区挖掘两阶段算法.第一阶段:从内聚的子团和度中心性节点的影响力出发,从网络中选取分散的k个内聚的且有影响力的极大完全图作为初始社区;第二阶段,基于局部社区模块度扩展方法,将重叠节点和初始社区外节点扩展到与其连接紧密的相应社区内.最后通过仿真实验验证了该算法的有效性.
Community mining is an important work in complex network analysis. There are a lot of good community mining algorithms, but most algorithms are based on the connection relationship between the nodes to discove cohesive social groups, but the nodes of actual network generally have different behaviors and influences. Based on this, we take full account of the characteristics of which nodes closely interconnected within the community and nodes that have different influences, propose a two-stage mining algorithm based on expansion of maximal-complete graph. The first stage： from the influences of the sub-group cohesion and degree centrality nodes, we select dispersed k cohesive and influential maximal-complete graphs as the initial communities： The second stage： based on local community module expansion method, the overlapping nodes and nodes outside initial communities are expanded to more closely connecting corresponding communities. Finally, experiment results show that the method is effective in detecting community structure.
出处
《河南科学》
2015年第12期2140-2145,共6页
Henan Science
基金
绥化学院科学技术项目(KQ1301007)
关键词
社区结构
极大完全图
度中心性节点
community structure： maximal-complete graph： degree centrality node
