复杂网络模糊重叠社区检测研究进展

摘要

模糊重叠社区检测通过扩展隶属度取值空间,实现了重叠节点与社区之间复杂且模糊隶属关系的精确化测量,不仅能够有效提升重叠社区结构检测的精确性,而且能够深度挖掘出节点和社区的重叠特性.文中首先分析了模糊重叠社区检测与传统离散重叠社区检测的关系;然后对二者的国内外相关研究现状进行阐述和分析,其中在模糊重叠社区检测方法研究中根据模糊隶属度获取方式的不同将当前相关研究分为扩展标签传播、非负矩阵分解、基于边界节点的两阶段检测、模糊聚类、模糊模块度优化五大类进行综述,重点分析了基于进化算法的模糊模块度优化方法;最后对模糊重叠社区检测研究未来的发展趋势进行了分析和展望.%Through expanding value space,fuzzy overlapping detection redefines the fuzzy membership degree,which can not only improve the detection accuracy of the complicated community structures,but also explore the overlapping features of nodes and communities.In this paper,we firstly give the explanation of the difference between crisp and fuzzy overlapping detection,and then summarize their related researches.To clearly state the fuzzy overlapping detection,we introduce the available work by dividing them into five classes on the acquisition method of fuzzy membership degree,including expanded label propagation,nonnegative matrix factorization,edge nodes based two-phase detection,fuzzy clustering and fuzzy modularity optimization.The advances and challenges of the fuzzy modularity optimization based on evolutionary algorithms are discussed in detail.At last some future research topics are given.