基于用户行为特征及关系的在线社交网络信息传播研究与建模

发布时间：2018-01-04 01:24

本文关键词：基于用户行为特征及关系的在线社交网络信息传播研究与建模　出处：《华东师范大学》2017年硕士论文　论文类型：学位论文

【摘要】：迅猛发展的在线社交网络已经成为人们获取和分享信息的重要平台,极大改变了人类社会的信息传播方式,在线社交网络上的信息传播也是近年来研究者们关注的热点。然而现有的研究成果虽然取得了一定成功,但依然存在以下问题:1)受限于实际数据的缺失,信息传播建模往往基于静态网络,依赖于用户接触和传播信息概率均等、用户行为相对静态等基本假设,难以描述真实信息传播中的复杂现象。2)随着近年来信息内容的复杂性和个体行为的多样性,用户具体通过哪些行为使信息得以传播、如何准确衡量用户的影响力和传播效率等问题,依然没有得到普遍认同的解释。围绕这些问题,本文利用大数据分析手段挖掘Twitter数据,基于用户行为特征的研究,对在线社交网络中信息传播过程进行了系统描述和动态建模分析。本文的主要研究成果和学术贡献如下:1)分析了在线社交网络中特有的用户行为特征,讨论了用户传播影响力的相关因素和评价方法。本文采集并利用了 Twitter平台的信息数据和相关用户数据,基于用户的信息转发行为构建了信息传播网络,分析了其网络拓扑结构性质和度分布,引入了基于用户发布或转发信息数量的用户活跃度指标,提出了基于等待时间和时间间隔的用户积极度和用户持续度指标,将转发数量作为传播影响力的评价指标,发现了用户活跃度难以和用户的信息传播影响力挂钩,而高用户积极度和持续度则是获得出色信息传播影响力的必要条件,发现了用户行为特征存在社区性和群体性差异。2)利用Fast-unfolding社团划分的方法,研究了在线社交网络中用户行为特征的社区性,解释了网络中的信息传播机制。基于Fast-unfolding算法对Twitter信息传播网络进行了社团划分,验证了用户行为具有社区特性。随后挖掘了用户的行为特征社区性的深层作用,发现了信息传播过程中的信息创造者、传播促进者、传播支持者和信息消费者这四类群体,并描述了他们的行为特征和先后发挥的创造话题、引发传播潮流、扩散传播以及终结传播的作用。3)提出了基于信息流行度和用户关系的动态信息传播模型。不同于传统的静态网络模型或仿真模型,本文提出了一种双层动态网络信息传播模型。通过随时间变化的发布信息构建了信息云网络,通过用户的关注关系构建了用户关系网络。综合考虑信息流行度、信息时效性和用户关系的影响,通过权重参数来衡量各因素的影响比例,动态计算了某个用户在某时对于某条信息的转发概率,实现了双层网络之间的动态联系,根据发布信息随时间的变化情况演化了信息转发的整个过程。最后,基于Twitter的信息发布、转发和用户关注数据检验了模型的可行性和结果的准确性,讨论了权重参数的最优取值,模型准确解释了流行度和用户关系在信息传播中的动态共同作用关系,可以用于在线社交网络的信息传播描述和预测。
[Abstract]:The rapid development of online social network has become an important platform for people to obtain and share information, greatly changed the human society of information dissemination, information dissemination focus on online social networks is also concerned by researchers in recent years. However, the existing research results have achieved some success, but there are still the following problems: 1) deletion limited by the actual data, the dissemination of information modeling are often based on the static network, depending on the user contact and dissemination of information equal probability, the basic hypothesis of user behavior is relatively static, it is difficult to describe the complex phenomena in the real.2 in information communication) with complex information content in recent years and individual behavior, through which the user specific behavior information spread, influence and spread efficiency issues such as how to accurately measure the user, still has not been generally accepted around these explanation. In this paper, using a large data analysis method of Twitter data mining, the research of user behavior feature based on the information dissemination process of online social networks is described and the dynamic modeling system. The main research results and contributions are as follows: 1) analyze the user behavior features of online social networks, discussed the related factors the user influence and evaluation methods. The acquisition and use of information data of the Twitter platform and the relevant user data, construct the information communication network based on the forwarding behavior of user information, analyzes the network topology properties and degree distribution, introduces the activity index of user number of user information published or forwarded based on the wait time and time interval of the user and the user continued positive degree index based on forward evaluation index number as the spread of the influence of the hair The user activity and information dissemination to hook the influence of users, and users of high product and duration is extremely necessary to obtain excellent information spreading influence, found the characteristics of user behavior in community and group differences of.2) by Fast-unfolding community division method, community study of user behavior in online social networks in the feature, explains the information dissemination mechanism in the network. The Fast-unfolding algorithm is partition of Twitter information transmission network based on the verified user behavior has the community characteristics. Then the effect of behavioral characteristics of deep mining community households, found in the process of information transmission and information dissemination of creators, promoters, supporters and communication these four types of information consumer groups, and describe their behavior characteristics and has the creativity of the topic, lead to the spread of the trend, the spread of the spread And the end of transmission of the.3) put forward the dynamic information model of information dissemination of popularity and user relationship. Based on static network model or simulation model is different from the traditional, this paper proposes a two-layer dynamic network information dissemination model. By changing with time release information to construct the information cloud network, through the user's attention to relationship building the user relationship network. Considering the popularity of information, information timeliness and customer relationship, through the weight parameters to measure the effect of each factor proportion, dynamic calculation of a user at a time for a forwarding probability information, realizes the dynamic relationship between the double network, according to the release of information changes with time the evolution of information forwarding of the whole process. Finally, the Twitter based information release, user data forwarding and attention to test the feasibility and accuracy of the results of the model The optimal selection of weight parameters is discussed. The model accurately explains the dynamic interaction between popularity and user relationship in information dissemination, and it can be used for information propagation description and prediction in online social networks.

【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：G206;TP393.09

【相似文献】