Research On Internet Public Opinion Propagation Mechanism Based On Multi-type Rumor Information

With the rapid development of mobile Internet technology,online socializing has gradually become the mainstream of people's life.Various social platforms make people's lives more and more convenient,but the ensuing Internet rumors have brought tons of damage to people.Therefore,research on the propagation mechanism of Internet rumors has become a hot issue nowadays.Studying the real trend of information propagation by using the rumor propagation model can not only deeply understand the network structure characteristics and user group behavior,but also have a good effect on the monitoring of network public opinion.This thesis takes multi-type rumor information as the entry point and studies the rumor propagation mechanism from the perspectives of network state and individual behavior.The core work is as follows:1.At the level of network state,considering that the existing models lack the consideration of the user's psychological game and the message function relationship under the joint propagation of multi-type rumor information,this thesis designs an information propagation dynamic model based on multi-type rumor information and dynamic games.First,the individual drive mechanism and the fan drive mechanism were used to extract the key factors affecting the spread of rumors from the inside and outside,and the regression method was used to measure the information influence.Second,this thesis introduced the dynamic game theory and formulated game strategies to deeply explore the competition and cooperation relationship among rumor,anti-rumor and prom-rumor(promoting romors)information.Last,this thesis constructed a SIAR model by adding an anti-rumor state in SIR model,and used this model to perceive the overall situation of rumor propagation.2.At the level of group behavior,in response to the problem that current researches on user behavior prediction mainly focus on explicit topology structure and ignore the influence of implicit relationship between users on rumor propagation,this thesis proposes a user behavior prediction model based on implicit link and GCN(Graph Convolutional Network).Firstly,the implicit relationship between users was mined by using KD-Tree(K-Dimensional Tree)to improve the network topology.Secondly,the GCN neural network model was used to complete the full representation of topic messages,network structure and user information.Meanwhile,this thesis completed the basic multiple classification by adding a Soft Max layer to the GCN model.Finally,combining with the idea of model fusion,this thesis constructed a predictive model of user forwarding behavior by using the voting mechanism.In addition,this thesis also used real data to experimentally verify the proposed method and model.Experiments have confirmed that the rumor propagation dynamic model can not only reveal the rumor,anti-rumor and prom-rumor information game propagation process more truly,but also can describe the propagation trend of multi-type rumor information in social networks better.The proposed user behavior prediction model can not only predict user forwarding behaviors under multi-type rumor topics effectively,but also more deeply excavate the driving factors of user behavior.The research of this thesis not only provides important theoretical basis for public opinion guidance,but also provides practical application value for public opinion management and control.