Research On Specific Information Diffusion Prediction And Restraint Techniques In Social Networks

In recent years,the amount of users and contents of social networks such as Twitter,Facebook,Sina Microblog,etc.has increased sharply.This is accompanied by more and more information that is difficult to distinguish the authenticity and needs to be suppressed.Nowadays information restrain techniques on online social networks has attracted growing interest among researchers.The existing researches are mainly based on identifying the authenticity of the information itself,cutting off the source of the communication,that is,grabbing the author;or removing some of the communicators in the network and cutting off the propagation path,but there are few researches focused on the suppression process.In view of the deficiency of existing researches,this thesis will start from the real Weibo social network.A specific information propagation path prediction model is proposed by user attributes and text data which models the user's points of interest and forwarding preferences.Then an information diffusion model is proposed to simulate the scene of information propagation.Select a set of seed users in the social network,inject them with immune information and spread them,so that they have the greatest influence on the spread in the social network and can quickly cover the user nodes that have been affected.The specific work of this thesis includes:Firstly,in order to predict the spread range of specific information,the dissemination mechanism and prediction method of information are studied in this thesis.This research turns this problem into a user forwarding prediction problem.The model takes into account the characteristics of users' forwarding behavior,and establishes the association between users and specific information based on the user's interest preferences,tag information,etc.,while the text content,user attributes,the user's recent Weibo content,the existing propagation path of the information,etc.are known.Then predict the forwarding of users,and return the forwarding probability.Furthermore,based on the prediction of the spread range of the information,this thesis designs the restrsaint models and mechanisms respectively based on the given coverage principles(for example,maximum coverage of nodes that are about to be spread by the specific information;maximum coverage of specific information status nodes and non-status nodes covering a certain proportion).A set of seed user nodes in the social network are found according to the seedset selecting algorithm in the model.Then inject immune information to reduce the spread of the original Weibo as much as possible to achieve the effect of inhibiting the spread of specific information.The social network specific information propagation suppression model proposed in this thesis is experimentally verified on a real social network dataset.Through experimental comparison and analysis,we can conclude that the social network specific information propagation suppression model proposed in this thesis can effectively abstract social networks.The process of information dissemination and suppression simulates the information dissemination steps in real scenes,and can effectively suppress the spread of bad information on social networks.Finally,the thesis obtains real data from the reports of the Weibo Community Management Center.Then capture the original weibo and their propagation path and select the users who participate in forwarding as seed nodes,and grab their followers and their weibo contents.On the basis,text preprocessing is carried out to reduce the irregularity of text contents.Based on the above work,we finally designed and implemented a social network specific information suppression system and visualized the dynamic propagation process of information.