AOGC: An improved gravity centrality based on an adaptive truncation radius and omni-channel paths for identifying key nodes in complex networks

Identifying key nodes that have a significant impact on network structure and function is a fundamental issue with numerous practical applications and technical challenges. The gravity model is a unique method for identifying key nodes that has piqued the interest of many researchers. Most of existing gravity-based methods have limitations in that they neglect the surrounding environment of a node in deeper level such as node location, neighborhood topology, higher-order neighbors, multiple interaction paths, and so on. They assume that inter-node interactions occur only along the shortest path, which is not true and renders them inaccurate in many cases. Based on our defined adaptive truncation radius and "looseness distance" considering omni-channel paths, we propose a new gravity centrality method that integrates multiple node properties and accurately characterizes node interaction distance. Numerical simulations indicate that (i) the gravity model is a good approximation of node interaction effect and (ii) a derivative method outperforms recent high-performing node importance evaluation methods in terms of distinguishing ability, accuracy, spreading ability, and imprecision function with an acceptable time complexity. Furthermore, the proposed method exhibits good stability on networks of varying scales and structural characteristics.