Modeling and identifying influential nodes in complex networks using graph embedding and temporal decay information

Identifying influential nodes in complex networks is a fundamental problem with important applications in information diffusion, epidemic control, and viral marketing. A major challenge in this task is that most existing centrality-based approaches rely primarily on static topological properties of networks while ignoring latent structural relationships and the temporal evolution of influence. In real-world social systems, however, the impact of nodes is shaped not only by their structural position but also by temporal dynamics such as the gradual decay of user attention to information over time. In this paper, we propose a novel framework for modeling and identifying influential nodes by integrating graph embedding techniques with temporal decay information (GETD). First, the network is mapped into a low-dimensional vector space using graph embedding, enabling the preservation of structural proximity and hidden relationships between nodes. Then, to better reflect realistic diffusion processes, we further incorporate a temporal decay mechanism that models the diminishing influence of nodes during information spreading. By combining embedding-based structural features with decay-aware dynamics, we design a hybrid influence estimation approach that captures both the network topology and the temporal attenuation of spreading processes. Extensive experiments conducted on multiple real-world networks demonstrate the effectiveness of the proposed GETD metric. Meanwhile, GETD achieves an average Kendall rank correlation of 0.852 and influence spread to a 2.4% improvement in node ranking performance.