Research On Information Diffusion Mechanism Of Social Hotspots Based On Topic Life Cycle

Current research on social hotspots propagation mechanism is mainly based on user topology space,and it focuses on micro propagation mechanism of single message and macro popularity prediction of multi-message.But in the real world,the generation of topics often includes a large number of interrelated network information and complex dynamic causes.And the topics also have some polymorphic characteristics of multi-dimension,multi-level and multi-domain in propagation space.At the same time,there are inherent relationships between the multiple key elements that promote topic propagation.This project takes the network topic information propagation mechanism as the research object,focusing on the multi-message mutual influence relationship,the polymorphism of the propagation space and the correlation of multi-key elements,and deeply discusses the information propagation mechanism of the complex network space.The launch of this study will reveal the internal mechanism of information dissemination in a deeper and more accurate way.This study takes the information dissemination mechanism of social network hot topics as the research object,takes the complete life cycle of information dissemination "past-present-future" as the main focus,and takes "information traceability-dissemination model-situation prediction"as the research direction,with more practical and in-depth research perspective,focusing on the interaction between multiple information of social topics,the multi-dimensional,multi-level,multi-domain,multi-stage and other characteristics of the communication space.The specific research interests include:key target traceability under multi-message iteration driven;multi-message,multi-domain and multi-role propagation dynamics mechanism in complex network space;multi-dimensional and multi-level group behavior trend prediction within social network hot topic.Through the project,we explore the quantization mechanism of multi-message interaction and the measurement method of multi-key target relevance in the multi-dimensional and multi-space social hotspots in the general form.In detail,the main innovations of this paper are mainly reflected in the following three aspects:1.In terms of information tracing,a model for relevance mining of multiple key elements in a topic-based multi-message dissemination network is proposed.The model divides the topic network into a multi-message dissemination network,builds a single message dissemination network through the forwarding relationship between users to determine the dissemination path;then,constructs a ternary correlation graph to illustrate the relationship between message nodes,path nodes and user nodes.Finally,an iterative scoring algorithm based on ternary association graph is proposed to mine three different types of key elements.The algorithm uses the initial influence score vector of different elements and the mutual weight matrix to obtain the final score sequence of various elements.The algorithm overcomes the concurrency and multi-path complexity of multiple messages under topics,and can effectively identify key elements.2.In terms of dissemination model,at the single-message hot topic level,a social hotspot dissemination dynamics model based on user dynamic interaction and evolutionary game is proposed.First,in real social networks,changes in hot trends will lead to dynamic changes in users'willingness to participate in hot topics.This effect is mainly reflected in the dynamic behavior of users.This paper uses evolutionary game theory to quantify the dynamic evolution process of users' willingness to participate in hot spots,and dynamically adjusts the infection rate of the information dissemination model according to the game relationship.Secondly,in view of the heterogeneity of the real network structure and the complexity of the heterogeneous mean field,evolutionary game is introduced to improve the heterogeneous mean field.On this basis,a new dynamic model of evolutionary game information dissemination is constructed.Finally,considering the dynamic behavior between nodes and the heterogeneity of real social networks,a heterogeneous mean field hotspot propagation model based on dynamic evolution mechanism and improved is obtained.At the level of multi-message hot topics,a dynamic model of information dissemination based on multi-message interaction and multi-space diffusion is proposed.First,in view of the multidimensionality of diffusion space,the overlap of network structure and the complexity of diffusion behavior,the layered mechanism is used to mine key features and map them to multiple diffusion spaces.Secondly,by constructing a multi-message diffusion path,we can further analyze and perceive the diffusion of each piece of information;at the same time,considering the dynamic change of the infection rate with time during the information dissemination process,we introduce time-dependent variables related to multi-message interaction,revealing The propagation process of mutual promotion(cooperative relationship)and constraint(competitive relationship)between multiple messages.Finally,considering the complexity of information interaction and diffusion networks,we introduce a dynamic interaction mechanism based on multiple SIS expansion models,and use the discrete-time micro-Markov chain method(MMCA)to construct a unified multi-message coupling dynamic process frame.3.In terms of situation prediction,at the user interaction level,a social hotspot forwarding prediction model based on implicit data dynamic compensation is proposed.First,according to the characteristics of tensors in data space conversion and projection,we propose a tensor-based user interaction behavior mining mechanism to solve the problem of inaccurate interaction strength calculation caused by sparse user interaction data,and analyze the relationship of concern The impact on user interaction.Secondly,the time decay function is introduced when constructing the tensor,which solves the problem of user interaction behavior over time,dynamically fits the user's behavior,and further describes the evolution trajectory of user behavior in current social hot spots.Finally,using time slicing and topic life cycle discretization,a user forwarding prediction model is constructed based on the logistic regression model.At the network structure level,a prediction model of group behavior based on network structure and factor graphs on hot topics is proposed.First,considering the dynamics of social topic information dissemination,we propose a traditional structure-based information triple structure to analyze complex influencing factors.Secondly,according to the Markov nature of information diffusion,using the basic concepts and methods of random field theory,we propose a group behavior communication model.The model can not only extract the influence of different factors,but also explore the development trend of topics.In a word,the thesis focuses on the life cycle of social network hot topics,with "past-present-future" as the main line,and conducts the research direction of social network hot topics "information traceability-dissemination model-situation prediction".