User Attribute Information Inference Method Based On Graph Neural Network

In recent years,with the popularity of the Internet and mobile Internet,social networking platforms such as microblog,wechat,QQ,etc.have risen rapidly and developed rapidly.These social networking platforms provide convenient and effective communication channels for people,and gradually become an indispensable part of people's daily life and work,and also have a large number of user groups.These users will generate a large number of network data on social platform,including user attribute information,user social information,etc.,among which user attribute information includes user gender,age,region and other specific information,which can help platform managers better serve users and operation management.However,due to the privacy protection of some users,the user attribute information is often missing or incomplete,which will affect the work of platform managers.In order to solve this problem,researchers began to study the problem of user attribute inference in social networks.However,the current research methods are mainly based on the attribute information of nodes or the local network structure to solve the attribute inference problem,without considering the influence of the deep topology information in the complex network on the attribute information inference of users.In this paper,the method of user attribute information inference in social network is studied on the data of social network of microblog.First of all,several graph embedding models are used to capture the structural equivalence and structural homogeneity of nodes in social networks,and different similarity calculation methods are used to calculate the similarity between neighbor nodes after graph embedding,and neighbor weighting method is used to solve the user attribute inference problem.Then,this paper uses the graph convolution neural network to learn the deep topology information in the social network data.Two experiments are carried out here.The first one is to train the network data using only the node features by using the graph convolution network model.The second one not only uses the node features in the network,but also combines the vector features of each node obtained by the graph embedding model.Furthermore,this paper uses the graph attention network model,introduces attention mechanism on the basis of graph convolution neural network,and introduces adaptive weight in each convolution,so as to better learn the attribute information of neighbor nodes received by the target node.In the experiment,it is proved that the graph neural network,especially the graph convolution neural network and the graph attention network,can be used to infer the user attributes.Graph neural network is of great significance to the structure of social network and the mining of user information in social network.For social network user attribute inference,we need to make use of the advantages of graph neural network to better study it.