An improved gravity model for identifying influential nodes in complex networks considering asymmetric attraction effect

Identifying influential nodes in complex networks is a crucial and challenging research issue in network science. Most existing algorithms rely on static characteristics of networks and operate under the assumption that interactions between nodes are symmetric. However, the potential asymmetric interactions between nodes pairs are often overlooked in real-world networks. To address this gap, we propose the Asymmetric Gravity Model (AGM), which identifies influential nodes in complex networks by considering asymmetric attraction effects. The core idea of the AGM algorithm is that a node's influence is calculated by accumulating the attractive forces of its neighboring nodes within a specified influence distance. Specifically, by introducing a newly developed asymmetric attraction coefficient, we transform the traditional adjacency matrix into an asymmetric attraction matrix. The proposed algorithm more accurately captures the relative attraction relationship between node pairs within networks. Meanwhile, we synthesize all potential attraction paths to adaptively determine the influence distance of networks. Furthermore, extensive experimental results on nine real-world networks demonstrate that the AGM algorithm outperforms eight competitive, state-of-the-art algorithms in terms of ranking accuracy, effectiveness, uniqueness, and the ability to accurately evaluate top-ranked nodes.