Node Importance Evaluation Method for Heterogeneous Networks Based on Node Embedding

In reality, complex systems are often represented by networks, and heterogeneous networks are more effective in describing the interaction behaviors among various elements. Evaluating the importance of nodes in a heterogeneous network is beneficial for maintaining the stability of the network. This study proposes a node importance evaluation method, named TAGCN Auto-Encoder Comprehensive Influence (TAE-CI), for heterogeneous networks, which combines graph neural networks with centrality measures. The method uses Topology Adaptive Graph Convolutional Networks (TAGCN) to improve graph autoencoders, encode different semantic subgraphs, reconstruct adjacency matrices, and optimize reconstruction loss to obtain node embedding vectors. To obtain the comprehensive influence of the nodes, the node embedding vectors are incorporated into the topological potential function to calculate the global influence, which is then combined with the local influence. The proposed method is evaluated on three real network datasets using the Susceptible Infected Recovered (SIR) model, and the results demonstrate its efficacy in evaluating node importance.