Research On Node Influence Propagation Algorithm Based On Knowledge Graph

Knowledge Graph,as a new method of knowledge representation,aims to describe the entities and concepts that exist in the real world,as well as the relationship between these entities and concepts.The application of large-scale knowledge graph originated from the introduction of knowledge graph into the search engine in May 2012,which opened the research and application of large-scale knowledge graph.With the development of Internet,great changes have taken place in the information environment of mankind,which has a profound impact on the communication and propagation of information.As knowledge graph carries a lot of information,and with the continuous improvement of knowledge graph,a huge relationship network is gradually constructed,and massive data is produced.The research and exploration of the information value propagation based on knowledge graph will be one of the important issues.The main contents and improvements of this paper are as follows:First,we analyze the classic epidemic model,and find that the classic epidemic model is an abstract description of the information propagation,and does not analyze the information propagation of the network at a micro point of view.In this paper,a message propagation method based on graph model is applied to transfer the probability distribution state of nodes to adjacent nodes,and get the probability distribution state of each node.Secondly,in the perspective of network communication,we deeply analyze the real network propagation effect in the process of network communication.It is found that the propagation of actual network is often caused by a certain node,and it is propagated in the community first,and then spread among the communities.Therefore,in the process of message propagation,this paper takes the community into consideration to improve the existing message propagation.Finally,based on the above models,we designed the following experiments.On a given graph structure dataset,we use separately epidemic model in the community,epidemic model between the communities,the belief propagation in the community,and belief propagation model between the communities to analyze and compare.Through experimental comparison and analysis,we find that epidemic model based on community structure can be more consistent with the propagation mode of real network in macro angle,and the belief propagation mode based on community effect can make a more detailed description of the network propagation process at the micro angle.In addition,compared with the original two propagation methods,the community effect in this paper is more consistent with the propagation effect of the actual network,and is proved to have a certain availability and effectiveness.