MNEGC: an improved gravity centrality based on node multi-features and network embedding for identifying influential nodes in complex networks

Abstract Identifying influential nodes in complex networks is a highly regarded and challenging problem. The use of gravity models to identify influential nodes has sparked research interest among scholars. However, existing gravity models mostly consider only limited dimensions of nodes and the shortest distance between nodes, which often leads to inaccurate identification. To overcome this problem, we propose a gravity centrality based on node multi-features and network embedding (MNEGC) for identifying key nodes. Firstly, we define the third generalized energy based on the generalized matrix, simultaneously considering the neighborhood coreness and clustering coefficient of nodes, and combining these three metrics to use as the mass of the nodes. Secondly, the Node2vec algorithm is utilized to map the nodes into a low-dimensional vector space and calculate the Euclidean distance between nodes. Finally, the score of the nodes is calculated using the new gravity model. We conduct comparative experiments by comparing the MNEGC algorithm with four gravity models and five state-of-the-art algorithms on nine networks. The experimental results suggest that MNEGC excels in ranking accuracy, monotonicity, imprecision function and precision in identifying the top-10 nodes.