基于局部影响力的网络社团检测方法 常振超，陈鸿昶，刘阳，（国家数字交换系统工程技术研究中心，郑州 450002） 摘要：发现复杂网络中的社在社会网络，生物组织网络和在线计算机网络等具备十分重要的意义，由于这些网络呈现出规模大的特点，基于全局结构的社团检测方法往往不能够适用，而从局部信息角度出发存在诸多问题，如对节点初始位置敏感、拓扑信息难以有效利用和社团扩展方向无法有效控制等问题。本文聚焦于研究局部社团方法，首先，借助于一种半局部的节点中心性衡量指标，找到与给定节点邻近的最具影响力的节点集合，这些节点在网络中具备较快的信息传播能力；然后，从这些具备较大影响力的节点集合发去检测局部社团结构。对真实网络和计算机生成网络的实验表明，本文所提的方法具有更高的识别性能，且由于其算法复杂度较低，能够很好的适用于大规模网络处理。 关键词：网络；社团检测；局部信息；影响力节点 中图分类号：TP391 Community detection based on local influential nodes in networks Chang Zhen-chao,Chen Hong-chang,Liu Yang,Huang Rui-yang (National Digital Switching System Engineering Technological Research Center, Zhengzhou 450002, China) Abstract: Detecting communities is of great importance in computer science biology and sociology networks. There have been lots of methods to detect community. Methods using global information are unsuitable to detect the communities in large-scale networks as the structure of the whole network cannot be detected. Recently, community detection based on local information has attracted many researchers’ attention. But traditional local community detection methods have a number of limits, such as the detection results are sensitive to the position of source node. In this paper we proposed a community detection algorithm based on local influential nodes. Firstly, a set of influential nodes based on semi-local centrality measure is constructed which has a faster spreading speed of information. Then, we apply these influential nodes instead of the original source node for local community detection. Community detection from the local influential nodes which have a higher spreading rate than random nodes can be more accurate and stable. Experimental results on both LFR benchmark networks and real networks show that our method can well detect local community. As the computational complexity is very low, our method is effective to explore local community structure of large-scale networks. Key words： networks; Community detection; Local information; influential no