$K$-GCN for Identifying Key Nodes in Complex Networks

Abstract:

      Accurately identifying key nodes is essential for evaluating network robustness and controlling information propagation in complex network analysis. However, current research methods face limitations in applicability and accuracy. To address these challenges, this study introduces the $K$-GCN model, which integrates neighborhood $k$-shell distribution analysis with Graph Convolutional Network (GCN) technology to enhance key node identification in complex networks. The $K$-GCN model first leverages neighborhood $k$-shell distributions to calculate entropy values for each node, effectively quantifying node importance within the network. These entropy values are then used as key features within the GCN, which subsequently formulates intelligent strategies to maximize network connectivity disruption by removing a minimal set of nodes, thereby impacting the overall network architecture. Through iterative interactions with the environment, the GCN continuously refines its strategies, achieving precise identification of key nodes in the network. Unlike traditional methods, the $K$-GCN model not only captures local node features but also integrates the network structure and complex interrelations between neighboring nodes, significantly improving the accuracy and efficiency of key node identification. Experimental validation in multiple real-world network scenarios demonstrates that the $K$-GCN model outperforms existing methods.