基于节点尺度特征的重叠社区检测算法  

Overlapping community detection algorithm based on node scale feature

作　　者：张黎烁[1] 高继勋[1] 

机构地区：[1]河南工程学院计算机学院,河南郑州451191

出　　处：《河南理工大学学报（自然科学版）》2016年第5期706-712,共7页Journal of Henan Polytechnic University（Natural Science）

基　　金：河南省科技攻关项目(162102310081);河南省高等学校重点科研项目(15A520053)

摘　　要：针对社会网络中重叠社区检测问题,在节点尺度特征下量化社区结构,用这些特性更易界定社区划分。利用合理假设来量化节点尺度的期望值,基于节点描述符集和谱算法建立算法模型,从而提出一种重叠社区检测算法。该方法允许节点同时属于多个社区,在社区重叠时同样可行。通过计算验证,算法对于整体边缘密度都有效。在2类网络中实验的结果表明,该算法在重叠社区检测中性能稳定、准确性高,能适用于目标特定的社区概念。For the problem of overlapping community detection in social networks,community structures were quantified in terms of the node scale feature.With these properties,communities can be partitioned more easily.Using reasonable assumptions to quantify the expected value of the node scale,a model was established based on node descriptor set and spectral algorithm,and a community detection algorithm was proposed based on this model.The method allows that a node belongs to multiple communities,thus it＇s efficient for the overlapping community.After computational verification,this algorithm is effective for the overall edge density.Experimental results from two networks show that the algorithm has steady performance and higher precision in the overlapping community detection.It can be applied to communities with particular target concept.

关 键 词：重叠社区检测 社会网络 节点尺度 边缘描述符集 自我中心网 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]