.png){kind=icon}
Return
Multi-indicator resilience analysis method based on GCN for graph network
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2026.133542.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Graph network resilience analysis is critical for quantifying the stability of complex systems under structural perturbations. However, existing methods primarily rely on pure topological metrics or domain-specific heuristics, significantly limiting their generalizability. To address this limitation, we propose a Multi-Indicator Resilience Analysis method based on Graph Convolutional Networks (MIRA-GCN). To comprehensively capture both structural stability and functional retention, we design an evaluation system comprising four core indicators: embedded feature difference, feature weight fluctuation, prediction accuracy variation, and score variation. By adaptively fusing these indicators during the learning process, MIRA-GCN effectively captures graph functionality representations and quantifies performance degradation under topological perturbations. Extensive experiments on seven diverse graph datasets demonstrate that our method achieves high ranking consistency (up to ) with traditional methods. Furthermore, MIRA-GCN exhibits strong generalization across various network modularities and topologies, providing a robust, multi-dimensional tool for automated graph resilience evaluation.
Keywords:
Graph resilience
Graph Convolutional Networks
Multi-indicator evaluation
Topological perturbations
Network stability
Journal
IF:
6.5
Papers:
2.5W
Citations:
6.5W
Organization
No organization information available
Cited Papers
No cited papers available