Research On Social Topic Group Behavior Prediction Based On Motivation-Rumor

With the rapid development of the Internet and the popularization of social software,the rumors have broken through the limitations of their spread,which gave birth to a new form of expression,namely online rumor.Predicting group behavior in the spread of online rumor plays an important role in sensing changes in public opinion on the Internet.Most of the traditional research on online rumor is based on the rumor itself,without emphasizing the symbiosis and confrontation of anti-rumor and motivation-rumor.Therefore,how to effectively measure and analyze the multi-message interaction under the rumor topic,how to mine the potential relationship between users under the rumor topic,and then to predict the group user behavior are urgent problems to be researched and solved.Based on rumor topics in the social network,considering user attributes and the influence of three types of message "rumor,anti-rumor,motivation-rumor",we conducted predictions of network rumor retweet behavior and multi-message user interaction behavior under rumor topic.The main research work and contributions of this thesis are as follows:Aiming at the diversity of rumor topic messages,this thesis proposes a rumor topic group behavior prediction model based on representation learning and motivation-rumor messages.First,considering the unsupervised learning ability of the represention,different information space features can be reduced,unified,and densely expressed.In particularly,a new representation was designed for the potential structural feature of the rumor propagation network: Rumor2 vec.Then,considering the spreading process of rumor,anti-rumor,and motivation-rumor,the cooperative competition relationship based on evolutionary game theory is adopted to construct a new network topology of propagation relations.Finally,a dynamic,uniformly represent,rumor propagation group behavior prediction model based on Game-GCN(Evolutionary Game Theory Graph Convolutional Network)is proposed.Aiming at the problem of multi-message interaction of rumor topics,a rumor topic group behavior prediction model based on rumor and motivation-rumor multiple types message interaction was proposed.First,from the perspective of the internal driving mechanism of user behavior,individual user factors are extracted.Secondly,based on the influence of rumor and motivation-rumor on user participation behavior and the dynamics of multi-message interaction propagation,the structural feature of social relationships under the influence of rumor and motivation-rumor interaction are extracted.Finally,in view of the applicability of the prediction model to the data,a multi-model fusion is used to construct a user rumor and motivation-rumor message mutual participation prediction model.Finally,the proposed model and method are verified by Sina Weibo real data.Experiments show that the model can not only effectively predict the group behavior of user under the rumor topic,but also more truly reflect cooperation and competition between three messages in the process of propagation.