针对绝大多数社区发现算法都存在着网络节点仅隶属于一个社区的假设,引入谱图理论与粗糙集理论来分析复杂网络社区,提出一种用于网络重叠社区发现的粗糙谱聚类算法RSC,该算法用上下近似来刻画网络节点的社区归属,边界表示社区之间共享的节点,通过优化重叠社区结构模块度来实现重叠社区发现.通过3个不同类型真实网络的仿真实验,结果验证了该方法的可行性与有效性.
Given the fact that the vast majority of Algorithms for communities discovery assume that one network node belongs to only one community, spectral graph theory and rough set theory are introduced into analysis of community structures in complex networks, an algorithm RSC, which is used in discovering overlapping communities, is proposed. The basic idea of RSC is to describe commu- nities membership of network nodes with lower, and upper approximation, describe the network nodes shared by different communities with boundary, and to mine overlapping network communities by optimizing overlapping community modularity. Experimental results on 3 real networks from different domains indicated feasibility and validity of our approach.