Research On Social Network Node Multi-label Classification Technology

With the rapid development of the Internet,people have made great progress in communicating through the Internet,and also formed a new basic form of social networking.Social networks are the virtualization of communication between people in the real world.Especially in recent years,the rapid development of social networks(such as Facebook,Tencent,Weibo)has attracted scholars in many fields to mine and analyze social networks.data.Thereby promoting social network research on advertising,public services,marketing,academic exchanges,etc.As a method of data mining and analysis,classification problems are also applied to social networks.By classifying the nodes in the network and capturing the nodes' interests,hobbies,relationships or other possible characteristics through tags.Multi-label classification of social network nodes is the main research direction of this article.Nodes in social networks can represent a person in reality or an overall organization.Each node contains a large amount of data,such as text,pictures,audio,video.Most of the previous node classification problems were mainly node single-label classification,and then because of the increasing number of data and attributes of nodes in social networks,the research on node single-label classification has been difficult to meet the requirements of node label classification,so this article mainly focuses on node multi-label classification.There are two research directions for the multi-label classification problem of social network nodes: one is to classify nodes according to their own attributes,and the other direction is to classify nodes according to the connection information between nodes.The main work of this paper is based on the fusion of the node's own attributes and connection information,and a node multi-label classification method based on graph convolution neural network is proposed.The main idea of this method is to propose a model that extracts the node's own attributes and connects the information through the graph convolution network to complete the multi-label classification of social network nodes.The model is topology graph data with node auxiliary information,and then introduces the coupling degree principle In the convolution operation of the topological graph data for a certain node,the convolution parameters of the connected nodes are different,and the corresponding connection information is weighted and summed according to the different convolution parameters to realize the content information of the node Fusion with connection information and achieve further classification tasks.Finally,through the experimental comparison of two public data sets,the different algorithm models in the existing research field are compared and verified,proving the feasibility of the algorithm model proposed in this paper on the node multi-label classification task.