Critical Nodes Identification Algorithm Based on ResNet-CBAM

• Published in: IEEE networking letters Vol. 7; no. 2; pp. 103 - 107
• Main Authors: Li, Xujie, Shao, Fei, Sun, Ying, Li, Haotian, Huang, Jiayi
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.06.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Artificial intelligence; Attention mechanisms; CBAM; Complex networks; Complexity; Computational modeling; Data mining; Deep learning; Feature extraction; Machine learning; Networks; Nodes; Redundancy; Resource allocation; Telecommunications; Training; Vectors
• ISSN: 2576-3156; 2576-3156
• DOI: 10.1109/LNET.2025.3572513

The identification of critical nodes in networks is of substantial practical significance. For instance, it can expedite information propagation within networks, target vulnerable links to enhance robustness, and optimize resource allocation by reducing redundancy and lowering costs. To improve the accuracy of critical node identification, we propose an algorithm that integrates complex networks, propagation models, and deep learning techniques. The algorithm generates low-complexity features that include the characteristics of nodes and their neighboring nodes. A ResNet-CBAM network is then designed to identify critical nodes. To assess node importance, a method has been proposed that considers both propagation range and propagation efficiency, using their product as the evaluation criterion. Experimental results show that, compared to various centrality-based algorithms and other deep learning methods, our proposed algorithm outperforms others in terms of recognition accuracy across different types of networks.