Seed node selection algorithm based on augmented graph embedding

Identifying influential individuals in information dissemination is an important topic in social network research; however, most studies have difficulty in effectively identifying influential nodes to maximize specific information on the network when the graph embedding representation is not rich and fine enough due to the complex network structure; to address the above problem, a seed node selection is needed. is not rich and fine enough due to the complex network structure; to address the above problem, a seed node selection algorithm model based on enhanced graph embedding is proposed. First, to enrich the graph embedding representation data, an embedding model combining breadth and depth-first search is First, to enrich the graph embedding representation data, an embedding model combining breadth and depth-first search is used for vector representation of node features. Second, to obtain a more refined embedding representation, a feature fusion gate mechanism is proposed to adaptively fuse the node embedding components learned by the graph neural network using the gate mechanism. Experiments conducted on Wiki-2 and the large dataset Epinions show that the proposed method outperforms other existing state-of-the-art methods in terms of the expected impact range metrics.