在线网络社会行为特征研究及其应用

发布时间：2018-03-14 19:57

本文选题：网络社区　切入点：正文抽取　出处：《南京航空航天大学》2014年博士论文　论文类型：学位论文

【摘要】：近年来,网络社区日益普及,其受到越来越多的人的关注。网络社区可以被看作一种信息系统,其中任何人都可以发表内容表达自己的观点,例如BBS、社交网络以及微博。网络社区的最基本特征是用户与用户之间的自由互动,其广泛发展开辟了很多新的研究问题:1)用户的交互数据被定义和嵌入在预先定义的数据结构中(这些数据结构是由网络社区站点自身提供的),如何完整准确的提取用户的交互信息;2)由网络社区衍生出的交互网络的拓扑特征如何?以及它们的演化模型和社区结构如何?3)网络社区中的社会交互既包含积极关系(好友关系),又包含消极关系(敌对关系),基于一个给定的话题,如何预测一个用户对另一个用户的当前态度;4)网络社区中,一些用户使用多个用户名或将自己伪装起来(通常被称为“网络马甲”)与其它用户沟通,如何能自动识别这些“网络马甲”?本文根据网络社区中复杂的用户交互信息,构建相应的用户交互网络,并研究交互网络的拓扑性质、演化模型、社区结构及网络中用户间的当前态度预测,在此基础上,提出一种自动检测“网络马甲”的算法。其主要创新点如下:1)提出一种网页内容抽取方法,该方法采用模糊关联规则标记网页正文行,利用滑动窗口技术选择正文块,可以有效抽取网络社区的用户交互数据;2)构建兴趣网络、语义网络及相似观点网络模型,进而提出一种基于“最近更新时间”概念的网络演化模型,可以有效模拟网络社区的“小世界”和“无尺度”特性;3)提出了一种基于博弈论的用户交互行为预测方法,并据此给出了两种用户在线社会属性的定义,可对网络社区用户进行有效分类;4)设计了一种并行模块度优化算法来挖掘网络社区中潜在社区,并结合用户身份识别和网络社区划分技术,提出了一种“网络马甲”自动识别方法。
[Abstract]:In recent years, the network community is becoming more and more popular, and it has attracted more and more people's attention. The network community can be regarded as an information system, in which anyone can publish their views. For example, BBS, social networks and Weibo. The most basic feature of the online community is the free interaction between users. Its extensive development has opened up a number of new research questions: 1) user interaction data is defined and embedded in predefined data structures (these data structures are provided by the web community site itself), and how can they be fully and accurately described. What are the topological features of the interactive network derived from the network community? What are their evolutionary models and community structures? 3) Social interactions in the online community include both positive and negative relationships (hostile relationships, how to predict a user's current attitude towards another user based on a given topic). How can some users, using multiple user names or masquerading themselves (often referred to as "online vests"), communicate with other users, and how can they automatically identify these "online vests"? According to the complex user interaction information in the network community, this paper constructs the corresponding user interaction network, and studies the topological properties, evolution model, community structure and the current attitude prediction of the users in the network. This paper presents an algorithm for automatic detection of "network vest". The main innovation of the algorithm is as follows: 1) A method of web page content extraction is proposed. The method uses fuzzy association rules to mark the text line of a web page, and uses sliding window technology to select a text block. We can effectively extract user interaction data from network community to construct interest network, semantic network and similar viewpoint network model, and then propose a network evolution model based on the concept of "most recent update time". This paper presents a prediction method of user interaction behavior based on game theory and gives the definitions of two kinds of online social attributes of users, which can effectively simulate the "small world" and "scale-free" characteristics of the network community. This paper designs a parallel modularity optimization algorithm to mine the potential communities in the network community, and combines the technology of user identification and network community division. This paper presents an automatic identification method of "network vest".
【学位授予单位】：南京航空航天大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP393.09

【相似文献】