Modified Affinity Propagation for Key Node Identification Based on Intra-Class Distances

Identifying key nodes is vital to the research of information control in network science and multiagent systems. This paper presents an enhanced affinity propagation clustering algorithm designed to identify critical network nodes by leveraging local and global information. Specifically, we introduce a novel influence matrix, which serves as a standard representation of node relationships. This matrix considers factors such as the number of neighbors and the number of shortest paths, providing a comprehensive view of node importance within the network. The proposed algorithm combines matrix iteration with evaluating local and global network structures to identify key nodes. And intra-class distance is introduced to pinpoint a smaller group of key nodes in large-scale networks. The experimental results on six real networks of different sizes show the validity of identification using the Susceptible-Infected (SI) model and network efficiency.