Abstract: Community detection aims to uncover the intrinsic community structure embedded in complex networks, which is a critical task in complex network analysis. In order to solve the node clustering selection problem that arises in clustering algorithm when the node similarity values are the same, this paper introduces a gravity model that can integrate node degree and effective distance, so that the similarity between identical node pairs asymmetric, thereby enabling the characterization of inter-node similarity at a finer granularity. The algorithm proceeds in two stages: first, it leverages node similarity to derive the associated communities; then, it merges these communities by modularity maximization to yield the final network’s community structure. Through comparative experiments, it is found that the proposed algorithm can effectively identifies communities in both artificially generated networks and real-world networks, and it achieves superior accuracy compared with other state-of-the-art algorithms of the same category.