Research On Information Propagation Mechanism Based On Graph Kernels And Graph Neural Networks

The rapid development of internet technology has made online social media gradually surpass traditional media and become the mainstream information acquisition platform.With the emergence of new social media applications such as Weibo,We Chat and Tiktok,people can use social media more quickly and conveniently to publish and obtain information.However,due to the lack of the review mechanism of traditional media,the threshold for users to release information is reduced,resulting in a large number of unauthenticated false information flooding in the network,causing serious network public opinion.Therefore,the research on information dissemination,especially the mechanism of false information dissemination,is of great significance.The current research studies the mechanism of information dissemination cascade from the perspective of empirical analysis and modeling research,and extracts features from the content,users and propagation structure of information to detect false information.A large number of empirical analysis shows that the propagation structure of information can well display the dissemination mechanism of information,and significantly distinguish between false information and real information.Therefore,the empirical research on the information propagation structure on social media can provide support for false information detection,which has important research significance and application value.Statistical features have limitations in capturing the global structure,while graph kernels can capture the sub structure of graph structure,and the graph neural networks can integrate the structure and attribute features of graph to embed graph expression.They have achieved good results in downstream tasks such as classification and prediction.Therefore,this paper intends to use the method based on graph kernels and graph neural networks to study the dissemination process of information on network social media.The main research contents and innovations are as follows:(1)This paper proposes an analysis method based on Weisfeiler-Lehman graph kernel to study the topological characteristics of information propagation structure,and empirically analyzes the structural differences between false information and real information in the process of dissemination and evolution.The results show that the evolution of information propagation structure has the trend of changing from full broadcast to multi broadcast,and the medium-sized true and false information propagation structure has the largest difference,and there are also significant differences in the propagation structure of true and false information in the early stage of transmission.(2)This paper proposes an analysis method based on graph auto encoders to express the characteristics of information propagation structure,and cluster the embedded expression to analyze the propagation difference between false information and real information.The results show that there are two completely opposite propagation modes in the information cascade.Most of the real information is highly concentrated in the broadcast based structure,while the false information has a multi chain structure.In the evolution process,the difference between the propagation structures of the true and false information in the medium-sized cascade is the largest.