Heterogeneous Network Node Classification Based on Graph Neural Networks

With the rapid development of modern science and technology, a large amount of data has been generated and accumulated in reality, forming heterogeneous networks with complex interaction relationships. Early data mining methods face problems such as insufficiently rich aggregated information when dealing with node classification in heterogeneous networks. To address this issue, this paper proposes a Heterogeneous Network Node Classification based on Graph Neural Networks model (HNNC-GNN), which utilizes the advantages of graph neural networks in extracting rich feature information to classify nodes in heterogeneous networks. In order to obtain rich semantic information in the heterogeneous network, the heterogeneous network is represented as a multi-layer network according to the node type and the multi-layer network is transformed into multiple semantic graphs based on meta-paths. For each semantic graph, Topology Adaptive Graph Convolutional Networks (TAGCN) is used for node feature aggregation and attention mechanisms is introduced to calculate the weights of each meta-path-based semantic graph and fuse different semantic features to obtain node embedding vectors. The node embedding vectors are input into a multi-layer perceptron for classification. Experimental results on three real network datasets demonstrate that HNNC-GNN can effectively extract semantic information from heterogeneous networks and improve node classification performance.