Research On Multi-Type Message Oriented Network Rumor Propagation Mechanism

With the rapid development of the Internet,people's means of accessing information has changed drastically.Internet rumors are also spreading throughout the social media platforms as news spreads.Because of their sudden and short-lived nature,online rumors have a great impact on people's fears.Therefore,the research on the mechanism of online rumor propagation has become one of the hotspots in academic circles,which provides theoretical support for public opinion departments to eliminate rumors and maintain the normal order of society by quantifying and analyzing the mechanism of rumor propagation and exploring the potential factors affecting rumor propagation.This thesis proposes the topic of social network rumors from the perspectives of micro user cognition and macro influence of multiple types of messages,and analyzes and models rumor message forwarding and macro message posture perception.The main work and contributions of this thesis are as follows:1.In terms of macroscopic multi-type message(Rumor & Anti-rumor &Stimulate-rumor)posture perception,this thesis proposes a model of information dissemination based on a tripartite cognitive game and multi-type messages.Firstly,for the polymorphic nature of user cognitive process under multi-type messages,the hyperactive state triggered by users' exposure to multi-type messages is proposed to describe users' mental state in the process of multi-type message cognition.Secondly,to address the contradictory and assimilation characteristics of the cognitive process,the three-way cognitive game theory is used to construct the user state transformation drivers.Finally,combining the multi-type message game with the polymorphism of users' cognitive process,a model of information dissemination based on three-party cognitive game and multi-type messages is proposed.2.In terms of microscopic multi-type message forwarding dynamics prediction,this thesis proposes the ICS-TCN rumor message forwarding dynamics prediction model.Firstly,considering the problem of user cognitive abstraction under multi-type messages,user cascade features are fused and extracted,and the textual information in the topic is characterized using a BERT pre-training model.Meanwhile,fused user features are obtained to enhance the prediction effect by social context vector.Secondly,considering the inefficiency of traditional hyperparameter search,an optimized cuckoo search method is proposed to optimize the hyperparameters of the prediction model.Finally,based on the above,an ICS-TCN rumor message forwarding trend prediction model is constructed,which enables certain rumors with potentially large influence range to be detected and monitored at the early stage of propagation.The experiments show that the message propagation model can not only realistically portray the user cognitive game process in the real microblogging dataset,but also effectively perceive the multi-type message propagation dynamics.At the same time,the ICS-TCN rumor forwarding trend prediction model can predict the actual rumor forwarding trend better.