Research On Node Influence Evaluation Method Based On Attention Graph Neural Network

At present,with the continuous expansion of the scale of Internet platform,the scale of virtual network social platform has been rapidly expanded.The rise and rapid development of social network has attracted the attention of researchers.Among them,node influence is a complex phenomenon recognized in social network,and has become a hot research topic in sociology and computer science.Compared with ordinary nodes,a few important nodes greatly affect the network structure and information dissemination,and because of their particularity in the network,they can affect the behavior of other individuals in a larger range.The research on the influence of nodes shows that the important nodes in the network can provide the research foundation for the research of the maximum influence of nodes and the research of recommendation ranking.Because of the huge scale of social network,noise nodes inevitably appear in the network.Noise nodes have false influence on the message transmission in social networks and some interference on the influence evaluation of normal nodes.The noise node detection and elimination can improve the quality of the network and evaluate the influence of the nodes more accurately.In this paper,the problem of node detection and influence evaluation of social network noise is analyzed.The sampling strategy method which integrates the node activity is proposed to improve the accuracy of the noise node detection model.A method of node influence evaluation based on attention graph neural network is proposed.The real social network data is used to test and analyze the model.This paper studies the detection of noise nodes and the evaluation of influence in social networks,combining the topology and attributes of social networks.Aiming at the noise node problem in social network,according to the user interaction characteristics and existing research in social network,the paper proposes a sampling strategy based on node activity,which can sample the node sequence,and improve the accuracy of the detection model.Considering the ability of node information processing in social network,considering the influence evaluation of social network nodes,integrating attention mechanism and mapping the output results reasonably,a method of node influence evaluation based on attention graph neural network is proposed to calculate and evaluate the influence of social network nodes.The feasibility and performance of the method are verified by comparing the real social network platform data.As the practical application research of the node influence evaluation method proposed in this paper,this paper uses microblog data set as the actual verification data.The comparison experiment shows that the model constructed in this paper evaluates the influence of nodes in the network,and the mean square error of nodes is smaller,and the relative ranking results are better.The model constructed in this paper has excellent comprehensive performance.At the same time,the classification performance of the model is tested by using Cora data.The experimental results show that the classification accuracy of the model is improved compared with other common models.