Finding influential nodes via graph embedding and hybrid centrality in complex networks

Finding influential nodes is essential for understanding the structure of complex networks and optimizing the dissemination of critical information. The key challenge lies in determining which nodes hold the most significance and how to identify and select a group of disseminators to maximize their influence. Therefore, researchers have proposed various approaches and centrality measures, each offering unique perspectives based on the network's topology. However, existing methods encounter inherent issues due to their sole consideration of node topology information. They also overlook the interconnectedness between nodes during the node filtering process, leading to imprecise evaluation results and limitations in terms of spread scale. In this paper, we introduce a novel scheme to tackle this problem in the context of social complex networks, termed graph embedding-based hybrid centrality (GEHC). Our proposed GEHC scheme starts by employing the DeepWalk graph embedding method to project the high-dimensional complex graph into a simpler, low-dimensional vector space. This mapping enables efficient calculation of the Euclidean distance between local pairs of nodes, allowing us to capture the proximity of nodes accurately. To further enhance the identification of influential nodes, we integrate network topology information and hybrid centrality indices. To evaluate the performance of our approach, we conduct extensive experiments on real-life networks using standard evaluation metrics. Experimental results on real-world networks demonstrate that our proposed scheme achieves a Kendall rank correlation coefficient close to 0.9, reflecting a strong correlation with the outcomes of the susceptible- infected-recovered model and validating its effectiveness in identifying influential nodes. The experimental results showcase the superiority of our approach inaccurately identifying nodes with high influence, surpassing the performance of traditional and recent methods in complex networks.