Research on Node Classification Based on Graph Convolutional Network and Graph Data Enhancement Technology
In graph convolutional networks,node classification is a basic problem,which involves label prediction of nodes in graphs.However,because graphs in the real world often have complex structures and noise,the accuracy of node classification is often unsatisfactory.In order to solve this problem,this paper proposes a method using graph neural network and graph data enhance-ment technology to improve the accuracy of node classification.First,we use graph data enhance-ment technology to preprocess the graph data,transform and extend the original training data to generate more training samples,so as to improve the generalization and robustness of the model.Then,we use the graph convolutional network model to classify the graph data nodes.Finally,we conduct several comparison experiments on the Cora data set.The experimental results show that using convolutional network and graph data enhancement technology can significantly improve the accuracy of node classification,and the accuracy of node classification on Cora data set is improved from 82.6％to 84.0％.
Graph convolutional networkGraph data enhancementNode classificationAccuracy rate
NETL
NSTL
万方数据
