社交网络群体情感行为关键问题研究

发布时间：2018-04-18 15:12

  本文选题：复杂网络 + 社交网络　； 参考：《北京邮电大学》2016年博士论文

【摘要】：随着Web2.0和移动互联网技术的快速发展和成熟,以Twitter、Facebook、新浪微博等为代表的社交网络已逐渐融入到人们的日常生活中。社交网络根据用户不同的功能诉求,将社会各个层次的特定用户聚集在一起,使得空间差异、时间差异等因素不再成为人们交流的障碍,实现了社会关系在虚拟网络上的延伸。随着社交网络的蓬勃发展,信息在社交网络中的传播变得更加普适和广泛,海量的用户可以方便的在社交网络中浏览新闻、关注热点、与好友或陌生人互动,每时每刻都可以通过评论、转发、发帖等行为表达自己多样的情感。虽然网络中的交流经常被认为是虚拟的,但情感会随着信息在网络中传播扩散,产生各种人与人之间的情感互动,甚至影响着网络用户在真实世界的行为表现。人类情感行为的研究一直以来都吸引着来自社会学、心理学、经济学、计算机科学等多个学科研究者们的兴趣,但由于人类情感的复杂性,研究者们也面临着种种挑战。社交网络中海量用户行为数据被实时记录,给了我们前所未有的研究人类情感行为的机会。社交网络用户群体的情感行为研究具有广泛的应用基础和重要的现实意义。本文以社交网络为研究对象,利用复杂网络理论和数据挖掘方法,对网络用户群体情感行为所涉及的若干关键问题进行初步探索和研究。主要研究内容包括:'用户分级别情感行为的分析建模仿真'、'基于多元情感的用户聚类分析'、'用户情感影响者发现模型'、'用户情感社团发现'。论文的主要工作和创新点如下:(1)基于新浪微博数据,提出了社交网络用户分级别情感发帖模型,并通过仿真验证了模型有效性。具体过程为:首先对用户微博内容情感进行分级,分析微博用户群体情感行为。分析发现社交网络用户群体在表达某一级别情感的发帖量均服从幂律分布,且幂指数随着情感级别趋向平和而增加,大部分用户通过微博表达情感时较为平和,需要表达激烈情感时,用户参与比例会减小。然后建立用户分级别情感发帖模型,该模型考虑了发帖用户受到周围情感环境因素的影响,以及自身情感的随机性变化。最后模型仿真验证了网络用户群体分级别情感发帖量服从幂律分布以及幂指数的变化趋势。(2)基于对社交网络用户多元情感行为的分析,提出了一种针对用户多元情感时间序列的相似性度量方法,并利用该方法对用户群体进行情感聚类分析。具体过程如下:首先利用多元情感词库提取出用户微博的多元情感向量,并构建多元情感时间序列用以描述用户情感行为。然后结合PCA相似性和距离相似性度量用户间的多元情感行为相似性,该度量既考虑了用户的情感波动,又考虑了情感表达强度。最后将该度量与经典的k-means聚类算法结合,提出多元情感聚类方法,并使用该方法发现不同用户情感群体,描述不同群体的情感行为特点。(3)基于社交网络的异质特点和网络用户间情感互动,提出了一种微博用户的情感影响者发现模型(EmotionRank)。具体过程如下:首先建立包含两种节点(用户、微博)和三种关系(转发、关注、发帖)的异质微博网络,然后利用微博情感相似性和用户多元情感行为相似性验证所构建网络的情感同配性,确认情感影响在该网络中存在。再利用两种相似性将该网络转化为只包含用户节点的同质网络,进而在网络中使用随机游走模型发现情感影响者。最后基于微博数据实验确认了该模型的有效性和优越性。(4)基于社交网络用户群体的情感同配性,可以确认网络用户会依据情感行为相似而链接聚集形成社团。本工作以社交网络拓扑结构为基础,提出构建了以关注用户间以及转发微博间的情感相似性为边权重的情感网络模型,再利用CNM和BGLL两种方法在用户情感网络中发现情感社团。为验证情感网络更适合发现情感社团,情感网络与利用其它三种网络节点相似性构建的三个无向有权网络以及一个无向无权网络进行了对比,情感网络与四个对比网络有着相同的网络拓扑结构和不同的边权重。对比实验结果表明利用情感网络所发现的社团内部用户之间的情感行为更加相似,用户间的转发微博有着更相近的情感。
[Abstract]:With the rapid development of Web2.0 and mobile Internet technology and mature, with Twitter, Facebook, Sina, micro-blog and other social networks have been gradually integrated into people's daily life. The social network according to the functional demands of different users, the specific users of all levels of society together, making the space differences, differences and other factors are no longer time people become obstacles in communication, realize the extension of social relations in the virtual network. With the rapid development of the social network, the spread of information in social networks become more pervasive and widespread, massive users can easily browse news in social networks focus, interact with friends or strangers, all the time through comment, forwarding, posting and other actions to express their feelings. Although the diversity in the network communication is often considered to be virtual, but the emotion with information on the net The network spread, emotional interaction between people and people, and even affect the performance of network users in the real world. The research of human emotional behavior has been drawn from sociology, psychology, economics, computer science and other disciplines researchers' interest, but because of the complexity of human emotions. The researchers also faced various challenges. Massive user behavior in social network data is recorded in real time, to study the behavior of the human emotions we hitherto unknown opportunities. Which has been widely used and important practical significance to research the emotional behavior of social network user groups. This paper takes the social network as the research object, using complex network theory and the data mining method, the preliminary exploration and Research on some key problems related to Internet users emotional behavior. The main contents include: users' Modeling and simulation analysis of 'level of emotional behavior,' analysis' multi user clustering based on emotion, emotional impact that users' model ',' user emotional associations found. Main work and innovations of the thesis are as follows: (1) Sina micro-blog data based on the proposed social network user level emotional post model. And the model is verified by simulation. The specific process is as follows: first, the classification of users of micro-blog micro-blog users group analysis of the emotional content, emotional behavior. The analysis found that the social network user groups are power-law distributions in the amount of post express a level of emotion, and the power exponent with the trend of peace and increase the emotional level, the majority of users micro-blog through the expression of feelings is relatively flat, need to express strong emotions, user participation ratio will be reduced. Then the establishment of user level emotional post model, the model considers the user posting Affected by the emotions surrounding environmental factors, as well as their own emotional random changes. Finally simulation model to verify the change trend of the network user group level emotion posting power-law and exponential. (2) analysis of the social network based on multi user emotional behavior, proposes a method for similarity measurement multi user emotional time series, and the emotion of the clustering analysis of user groups by using this method. The specific process is as follows: firstly, using multiple emotion lexicon extracted multiple emotion vector of micro-blog users, and build a multiple time series is used to describe user emotional emotional behavior. Then combined with PCA similarity and distance similarity measure multiple user emotional behavior the similarity between the measurement of both emotions of the user, and consider the emotional expression intensity. Finally the measure with the classical K-means clustering With the algorithm, put forward multiple emotion clustering method, and used this method to find different user groups emotion, the emotional behavior characteristics of different groups. (3) the emotional interaction between heterogeneous characteristics and network users based on social network, put forward an emotional impact of micro-blog users found model (EmotionRank). The specific process is as follows: first includes the establishment of two kinds of nodes (users, micro-blog) and three relations (forwarding, attention, post) heterogeneous micro-blog network, and then use the micro-blog emotional similarity and multiple user emotional behavior similarity verification construction network emotion homogamety, confirm the emotional effects present in the network. Then two kinds of similar will the network into a user node contains only homogeneous network, and then using the random walk model found that the emotional impact in the network. Finally, based on the experimental data of micro-blog confirmed the validity of the model And superiority. (4) social network user groups based on emotional homogamety, can confirm the network users will be based on emotional behavior is similar and link together to form a club. The work is based on the social network topology, put forward to concern among users and forwarding micro-blog emotion between similarity of emotional edge weight network model then, using CNM and BGLL two ways to find emotional associations in the emotion of the user network. In order to verify the emotional network more suitable for emotional associations, emotional network and the use of the other three kind of network node similarity constructed three undirected weighted network and an undirected unweighted network compared with emotional network the network topology of the same and different edge weights and four contrast network. Experimental results show that more similar emotional behavior between the community users have discovered by using the emotion of the network, with the The inter - Household forwarding micro-blog has a more similar emotion.

【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP393.09;TP391.1

【参考文献】

中国期刊全文数据库 前8条

1 方滨兴;贾焰;韩毅;;社交网络分析核心科学问题、研究现状及未来展望[J];中国科学院院刊;2015年02期

2 丁兆云;贾焰;周斌;唐府;;社交网络影响力研究综述[J];计算机科学;2014年01期

3 曹盼盼;阎春宁;;人类通信模式的幂律分布和Zipf定律[J];复杂系统与复杂性科学;2009年04期

4 李楠楠;张宁;周涛;;人类通信模式中基于时间统计的实证研究[J];复杂系统与复杂性科学;2008年03期

5 周涛;;在线电影点播中的人类动力学模式[J];复杂系统与复杂性科学;2008年01期

6 安世虎;都艺兵;曲吉林;;节点集重要性测度——综合法及其在知识共享网络中的应用[J];中国管理科学;2006年01期

7 陈勇,胡爱群,胡啸;通信网中节点重要性的评价方法[J];通信学报;2004年08期

8 李鹏翔,任玉晴,席酉民;网络节点(集)重要性的一种度量指标[J];系统工程;2004年04期

中国博士学位论文全文数据库 前1条

1 孙静;群体性事件的情感社会学分析[D];华东理工大学;2013年

中国硕士学位论文全文数据库 前1条

1 郑兰;微博客世界中用户间互动对用户微博使用行为的影响研究[D];北京邮电大学;2012年

，

本文编号：1768899