Modeling Of Information Propagation In Online Social Networks And Reposting Prediction

In recent years, with the rapid development of information technology and thewidespread use of intelligent communication tools, the online social networks (OSNs)has become an important way in sharing and spreading information. Especially for themicroblog platforms, which has millions of users and promotes the emergency of hugeinformation transmission networks. In addition, they also promote human society to entera self-media age and the innovation of business forms. The purpose of this paper is toreveal information propagation mechanisms and the inherent laws both microcosmicallyand macroscopically according to dynamic model and data mining methods. Taking theeffect discussion of network topology on the spread of information as the breakthroughpoint, we propose several dynamic models with three mechanisms, including superspreading, interest decay and social reinforcement mechanism, as well as cooperative andcompetitive dynamics model. Based the models above, we further discuss online socialnetwork information dissemination mechanism and evolution trend from the compositemechanism of single mechanism, composite mechanism and multi informationinteraction. Meanwhile, an interest weighted random forest model is proposed to improvethe prediction effect of repost behaviors, and differs from the importance of different userproperties to forwarding behaviors.This paper can promote the implementation of traditional dynamic model indescribing the online user characteristics, leading to effective integration of user behaviormodel and dynamics model. The research concludes the essential difference betweeninformation communication and the spread of disease, clearly revealing the the intrinsiceffects on information dissemination because of different user attributes. Relevantconclusions can not only accelerate the integration of micro user behaviors model andmacro network information propagation modeling, but also optimize the public opinionguidance strategy for government departments and offer reference to promote thebusiness mode innovation.The main contents of this thesis are listed as follows. Firstly, the relations among network topology and dissemination of informationbased on four social networking platform is discussed. We introduce several dataset ofonline social network, such as Twitter, LiveJournal, Sina and Tencent microblogs, andanalyze the network topology by the degree distribution, small-world and scale-freefeatures based on the the basic theory of complex networks and the online socialnetworks. We compared the relationship between the he network topology and the useraction fanilly.Secondly, this paper analyzes the influence of super spreading mechanism oninformation spreading in OSNs from the aspect of single mechanism. We describe thesuper spreading mechanism. According to the cases of Sina microblogs, this describes thedefinition of super spreading mechanism. Then the dynamics equations is proposed basedon SIR model, and the global stable point is obtained by stability theory. Finally, thechange trend of different network nodes density under the effect of super spreadingmechanism is analyzed using numerical simulation method.Thirdly, the interest decay and social reinforcement mechanism in online socialnetworks is further analyzed. Based on the latest research, the concepts of interest decayand social reinforcement mechanism are defined, which can be represented by non-linearfunctions. Then, we analyze the network transmission threshold by the change of nodesstates. In the end, by using numerical simulation method, we reveal the the law ofinformation spreading in OSNs under interest decay and social reinforcement mechanismwith the best effect of stifling probability.Fourthly, the cooperative and competitive dynamics model for informationpropagation is proposed from the aspect of information compositing mechanisms. Weintroduce the analytical framework of population dynamics, and propose an informationpropagation model based on velocity, which has been tested by the actual data from SinaWeibo. And then, we propose a Lotka-Volterra cooperative and competition propagationmodel, thus to expounding the mechanism of cooperation and competition existpropagation process of information network and working out its system equilibrium pointand its stability. A simulation analysis is given according to the the actual data from sinaWeibo. Finally, we analyze the information propagation problem form the perspective ofusers’ micro behavior by the method of machine learning. This paper analyzes thedistinguishing abilities of user characteristic attributes, behavior property, active degreeand interest attributes, and designes an index system for reposting prediction based onTencent database. By introducing the idea of feature weighting technique, we propose aninterest weighted random forest model. The empirical study analyzes the performance ofimproved random forest model, compares the effects of users’ characteristics, behavioralattributes and interest attributes imposed on repost behaviors, thus revealing theimportance of user attributes in promoting the information propagation in OSNs.