Graph Neural Network Algorithm Based on Graph Convolution and Attention Mechanism

With the wide application of graph neural network (GNN) in many fields, how to extract and aggregate node features effectively has become a hot research issue. In this paper, we propose a graph neural network algorithm that leverages graph convolution and the attention mechanism to enhance the expressive power and aggregation efficacy of node features within a single subgraph. The model's core framework comprises two primary steps: Firstly, node features are extracted from subgraphs utilizing the graph convolution algorithm. In particular, the graph convolution operation updates the feature representation of each node by disseminating information from its adjacent nodes. This process can capture local structural information and effectively integrate the neighborhood features of nodes, thus enhancing the semantic representation of features. Secondly, using the extracted node features, the model incorporates an attention mechanism to aggregate the features in a weighted manner. By evaluating the similarity between each node and its neighboring nodes, the attention mechanism is able to dynamically assign weights to various nodes, thereby emphasizing the significance of key nodes during the aggregation process. This method overcomes the limitation of treating all neighbor nodes equally in the traditional aggregation method, and can more accurately reflect the relevance and importance of the nodes in the subgraph. The experimental outcomes show that the proposed algorithm outperforms existing techniques in terms of performance across various graph datasets.