Research On Information Diffusion And Intervention In Online Social Networks

The online social networks,represented by Weibo,Twitter and Wechat,have the characteristics of wide audience,fast communication and large influence,becoming an important basic platform for information sharing and communication.However,the large-scale spread of fake and harmful information on this platform seriously not only affects people's daily lives,but also threaten social stability and national security.Research on the information dissemination model of online social networks provides the basis theory for forecasting the large-scale dissemination of information,the research of intervening technologies such as source nodes locating,and information diffusion blocking provide technical support for eliminating bad nodes and avoiding large-scale diffusion of information.However,there are still the following problems to be studied: 1)The epidemic model is well fitted for describing large-scale information dissemination,but for discribering social network information diffusion,its accuracy of fitting and forecasting is relatively low,and the model can not forecast the trend of large-scale information transmission accurately.2)When the number of information sources is uncertain,the accuracy of locating the sources is low relatively,and then can people not smash all the bad nodes.3)The current information dissemination blocking model has not considered the difference of the node influence,and the blocking target is not accurate,and could block the bad information effectively.In view of the above problems,relying on the military research project,we carry on the specialized research to the online social network information dissemination and intervention technology.In this paper,firstly we propose a new discrete sequence propagation model based on the principle that using the k-core structure as the propagation unit instead of the random infection of the node,building the infectious disease model.Then,for locating uncertain number source nods,the time and space similarity between the network node and the information source is researched,determining the number and location of the source nodes.Finally,to effectively block information diffusion,the nodes' influence is added to establish a new blocking model,increasing the effectiveness of transmission blocking.The main research contents include:1.A k-core based information dissemination model is proposed.Firstly,the k-core network structure is used as the basic transmission unit,and the random encounter mechanism of the infectious disease model is replaced by the independent cascade.Then,the number of the information transmission nodes is computed based on the connection of the k-core's internal structure,and the propagation range is predicted according to inter-core connection relationships.Finally,the discrete time series and intra-kernel slow release mechanism are mixed to fit the propagation process and predict the propagation trend.The simulation results show that the accuracy of the fitting is slightly better than that of the SpikeM model,and the prediction error is about 15% lower than that of traditional methods.2.A method of traceability based on time and space similarity is proposed.Firstly,according to the mapping between the the probe nodes receiving time and the information propagation paths,the probe nodes that receive messages many times are analyzed for the source nodes' directivity,and the random walk algorithm is used to select the source selection.Then,the time similarity between the single receiving source nodes and the option source nodes are analyzed,and the source node localization problem is transformed into the clustering problem.Finally,clustered by the time similarity,the improved affinity propagation algorithm determines the number and location of source nodes.The simulation results show that the accuracy of the method is 15% better than that of the traditional methods,and the locating error is controlled within 3 hips.3.A node influence-oriented blocking model is proposed.Firstly,considering the influence of the nodes in the network,the blocking function model is established,and the objective function is to minimize the influence of the nodes receiving the information by removing the optimal node set.Then,the blocking model is analyzed that greedy algorithm could not guarantee the optimum solution,and a method based on sampling average approximation is proposed,the stochastic optimization problem of is transformed into the deterministic optimization problem,and further coded as a mixed integer programming problem.Finally,an improved quantum genetic algorithm is proposed to select the optimal block nodes.The simulation results show that the model proposed is better than the traditional blocking models.