邻域粗糙化的启发式重叠社区扩张方法

发布时间：2018-05-01 13:16

本文选题：粗糙集理论 + 邻域粗糙化　；参考：《计算机学报》2013年10期

【摘要】：重叠社区发现研究是当前图挖掘领域的前沿热点.基于结构适应度的局部扩张方法是其中一类可同时探测社区重叠和层次结构的方法.该文首先提出了基于邻域粗糙化的社区局部扩张方法,然后给出了一种反映社区内在结构特征的稳定性度量.针对局部扩张方法计算冗余和社区漂移等不足,采用一种新的种子社区启发策略来降低复杂计算和提高探测能力.在结构适应度最大化的条件下,以极大度节点的团作为种子社区进行局部扩张,通过社区稳定度度量对近邻重复社区进行合并,生成自然重叠的社区.在真实网络上的实验结果表明邻域粗糙化的方法可以有效地发现重叠社区,并具有很好的扩展性.
[Abstract]:Overlapping community discovery is a hot topic in the field of map mining. The local extension method based on structural fitness is one of the methods to detect community overlap and hierarchy simultaneously. In this paper, a local community extension method based on neighborhood roughness is proposed, and then a measure of stability reflecting the structural characteristics in the community is given. A new seed community heuristic strategy is proposed to reduce the complexity of computation and improve the detection ability in view of the shortcomings of local expansion method such as computational redundancy and community drift. Under the condition of maximum structural fitness, the cluster of maximal nodes is used as the seed community for local expansion, and the adjacent repeated communities are merged by the measure of community stability, and the overlapped communities are generated. The experimental results on real networks show that the method of neighborhood roughness can find overlapped communities effectively and has good expansibility.
【作者单位】：同济大学计算机科学与技术系;同济大学嵌入式系统与服务计算教育部重点实验室;太原理工大学计算机科学与技术学院;
【基金】：国家自然科学基金(61273304,61202170,61103067,61075056) 教育部博士点优先资助项目中央高校基本科研业务费专项资金资助~~
【分类号】：TP393.0

【参考文献】