面向中文微博的社区发现和结构洞挖掘的研究

发布时间：2018-03-11 00:36

本文选题：微博　切入点：社会网络　出处：《哈尔滨工业大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着互联网技术的快速发展，虚拟社会网络逐渐成为了人们关注的焦点。人们可以在虚拟社会平台中更加方便地认识好友和相互交流，从而促进了真实社会关系的发展，使得人类的活动与互联网的应用融为一体。而微博作为一种新兴的网络信息交互平台，其优越独特的设计理念使得用户之间的交流更加及时、方便与畅通，庞大的用户群体以及海量的文本信息蕴含着极大的价值，在不同的研究领域有着重大的意义。本文对微博网络中的社区发现和结构洞挖掘的主要研究如下：首先，阐述了社会网络的基本概念及主要特征，对网络分析的相关理论和技术进行分析，即文本的预处理、文本的表示模型以及常用的聚类算法等，同时，对文本的语义分析理论进行研究，了解特征空间降维的必要性以及隐含语义分析的基本流程。其次，分析微博网络的主要特点及其交流机制，并采用微博开放平台实现了数据的获取。同时，微博网络呈现出多维度的结构特征，而传统平面结构的网络模型不能体现不同主题领域下的权威度，因此，提出了基于主题的多层次网络模型，从不同的层次建立不同主题的用户交互网络图，以充分挖掘出网络中的社会群体。再次，分析微博网络的社区特征，并在传统的社区发现算法基础上，考虑到社区的主题特性，采用了词共现度的事件划分以及隐含语义的进展分析，实现基于主题相似性的社区发现。同时，从社区的网络结构出发，提出了基于最大基数搜索的Greedy算法以及基于权重平衡的WEBA算法，实现了网络社区结构的合理划分。最后，阐述社会网络结构洞节点的基本概念，以及其在社会资产以及不同领域中的重要作用，通过PageRank算法对网络节点的重要度做出了重要评价。并在此基础上，从网络的社区结构和流网络的信息扩散角度出发，提出基于舆论领袖的HIS算法和基于最小割集的MaxD算法，实现了结构洞节点的挖掘。
[Abstract]:With the rapid development of Internet technology, virtual social network has gradually become the focus of attention. People can more conveniently know friends and communicate with each other in the virtual social platform, thus promoting the development of real social relations. As a new network information exchange platform, Weibo's superior and unique design concept makes the communication between users more timely, convenient and smooth. The huge user group and the massive text information contain the great value, has the great significance in the different research domain. This article mainly studies the community discovery and the structure hole mining in Weibo network as follows:. Firstly, the basic concept and main characteristics of social network are introduced, and the related theories and techniques of network analysis are analyzed, such as text preprocessing, text representation model and clustering algorithm, etc. The semantic analysis theory of text is studied to understand the necessity of dimensionality reduction in feature space and the basic process of implicit semantic analysis. Secondly, it analyzes the main characteristics of Weibo's network and its communication mechanism, and realizes the acquisition of data by use of Weibo's open platform. At the same time, Weibo's network presents multi-dimensional structural characteristics. However, the traditional planar network model can not reflect the authority degree under different subject areas. Therefore, a multi-level network model based on topic is proposed to build user interaction network graph with different topics from different levels. In order to fully excavate the social groups in the network. Thirdly, it analyzes the community characteristics of Weibo network, and on the basis of the traditional community discovery algorithm, taking into account the thematic characteristics of the community, uses the word co-occurrence degree of event division and the implicit semantic progress analysis. At the same time, based on the community network structure, the Greedy algorithm based on maximum cardinality search and the WEBA algorithm based on weight balance are proposed to realize the reasonable division of the network community structure. Finally, the basic concept of the hole node of social network structure and its important role in social assets and different fields are expounded. The importance degree of network node is evaluated by PageRank algorithm. From the point of view of the community structure of the network and the information diffusion of the flow network, the HIS algorithm based on the opinion leader and the MaxD algorithm based on the minimum cut set are proposed to realize the mining of the structure hole nodes.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092

【相似文献】