作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2018, Vol. 44 ›› Issue (11): 178-183. doi: 10.19678/j.issn.1000-3428.0048583

• 人工智能及识别技术 • 上一篇    下一篇

基于网络拓扑与节点元数据的社团检测算法

刘宇廷,毕海滨,郭强,倪颖杰   

  1. 江南计算技术研究所,江苏 无锡 214000
  • 收稿日期:2017-09-07 出版日期:2018-11-15 发布日期:2018-11-15
  • 作者简介:刘宇廷(1987—),男,硕士研究生,主研方向为机器学习;毕海滨,助理工程师;郭强,博士;倪颖杰,高级工程师、博士
  • 基金资助:

    国家自然科学基金(91430214);“核高基”重大专项(2015ZX01040-201)

Community Detection Algorithm Based on Network Topology and Node Metadata

LIU Yuting,BI Haibin,GUO Qiang,NI Yingjie   

  1. Jiangnan Institute of Computing Technology,Wuxi,Jiangsu 214000,China
  • Received:2017-09-07 Online:2018-11-15 Published:2018-11-15

摘要:

传统社团检测算法利用网络拓扑挖掘社团结构,忽略了真实复杂网络中节点自身属性等信息在社团归属方面的重要作用。为此,提出基于网络拓扑与节点元数据的复杂网络社团检测算法。将高维的节点元数据建模为混合高斯模型,结合随机块模型建立似然概率模型,通过求解模型最优解得到网络的最优划分结果。在基准网络与Facebook网络上的实验结果表明,该算法不仅能准确挖掘网络中的社团结构,而且可结合真实社团情况给出合理解释

关键词: 复杂网络, 社团检测, 节点元数据, 高斯混合模型, 随机块模型

Abstract:

Traditional community detection algorithms mine the community structure by using the topology of networks,and due to the neglection of the attributes of the nodes in the real-world networks,the structure and meaning of the community can not interpret well.In view of this,this paper proposes the Community Detection algorithm based on Network Topology and Node Metadata(CDNTNM) for complex network.In this algorithm,the high-dimensional node metadata is modeled as a Gaussian Mixture Model(GMM),and the probabilistic likelihood model is constructed by combining the network topology model and GMM.The optimal devision of the network is obtained by solving the optimal solution of the model.Experimental results on the benchmark network and the Facebook ego-networks show that the proposed algorithm can not only excavate the community structure in the networks,but also give a more reasonable explanation.

Key words: complex network, community detection, node metadata, Gaussian Mixture Model(GMM), random block model

中图分类号: