项目名称: 节点内容和链接相结合的大规模内容网络社区发现方法及应用研究
项目编号: No.61473030
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 其他
项目作者: 贾彩燕
作者单位: 北京交通大学
项目金额: 80万元
中文摘要: 现有的复杂网络社区发现方法大都基于社区结构的原始约束:社区内的节点链接稠密而社区间的节点链接稀疏来检测网络中蕴含的社区,忽略了节点上的属性信息。而现实世界网络节点上常含有丰富的属性信息,如通信网络中的用户信息、通话位置、通话时长等,且两个节点可在属性上非常相似,但这两个节点间却不存在链接关系。如何同时利用网络节点间的链接关系和节点上附着的属性信息,准确、高效地发现网络、特别是大规模网络中的社区结构越来越受到人们的重视。本项目主要研究节点内容和链接相结合的高效社区发现方法及相关方法在微博网络用户群及话题识别中的应用。具体内容包括:1.基于概率模型节点内容和链接相结合的高效社区发现方法;2.基于非负矩阵分解节点内容和链接相结合的新型社区发现方法;3.基于采样策略和网络稀疏化表示的大规模网络节点内容和链接相结合的高可扩展性社区发现方法;4.相关方法在微博网用户群及话题分析等内容网络中的应用。
中文关键词: 数据挖掘;社会网络分析;社区发现;聚类分析;大规模内容网络
英文摘要: Community structure is one of the most essential features of real networks. Previous study mainly focused on identifying communities in networks using only links (relationship between pairs of nodes), or clustering nodes by content (features) of nodes. It is still a problem on how to make use of content of nodes and links together to detect communities in content networks high efficiently. In this project, we intended to study methods on community detection combining content and links in content networks, especially, massive content networks, present new methods and models to extract user groups and topics in Weibo networks. In summary, we will concentrate on the following studies. 1. Study on high efficent probabilistic approaches for detecting communities in networks by combining content of nodes and links together. 2. Study on NMF (Nonnegative Matrix Factorization) models for detecting communities in networks using content of nodes and links. 3. Study on high scalability community detection approaches based on sampling strategies and network sparsification on massive content networks. 4. Use the proposed models and methods to extract user groups and topics on Weibo networks constructed by Weibo users, users' followers, Weibo texts and texts' followers, and to discover communities in other content networks.
英文关键词: Data Mining;Social Network Analysis;Community Detection;Clustering;Massive Content Networks