Identifying Key Nodes Based on Neighborhood Topology and Voting Mechanism in Complex Networks

Large-scale networks cannot be effectively addressed by global structure-based techniques due to their high temporal complexity, while local structure-based methods may overlook global information. To overcome these limitations, we propose a novel key node identification method for complex networks, named cycle structure, voting mechanism, ranking principle (CVR). This method adopts a multilevel processing approach and an enhanced voting mechanism. Initially, it incorporates the centrality of the network cycle structure and describes the topological locations of nodes within their neighborhoods. Subsequently, the traditional voting mechanism is refined by incorporating both global and local information from complex networks, providing a more accurate representation of relationships between nodes and the structures of neighborhoods in the network. The extended neighborhood ideology is then integrated with the improved voting mechanism, resulting in an effective method for identifying hidden key nodes. The effectiveness of the CVR method is validated through experiments on nine datasets using nine baseline methods, including the susceptible, infective, recovered (SIR) and linear threshold (LT) models, as well as experiments involving the seed selection technique for choosing initial infection nodes. Results show that CVR improves the infection rate by 4.7%–156.8% under varying infection probabilities in the SIR model.