Heterogeneous graph neural network model based on attention and convolution aggregation

In the face of the rapid expansion of heterogeneous information networks, extracting the characteristics of multi-type nodes and edges effectively has become the core issue of graph neural network research. This paper innovatively proposes a model named HGATCA (Heterogeneous Graph Neural Network with Attention and Convolutional Aggregation), specifically for the problem of feature integration and information dissemination in heterogeneous maps. The model takes the lead in using convolution operations to deeply learn the features of nodes under different meta paths, which significantly enhances the efficiency of information extraction in specific contexts. Then, HGATCA cleverly integrates the attention aggregation strategy, delicately integrates multiple semantic information, and extracts the high-quality semantic features after fusion. Through a series of empirical tests on publicly available heterogeneous graph datasets, HGATCA has demonstrated superior performance over existing mainstream models on key tasks such as node classification and clustering.