HR-based: Structure-preserving link prediction through hierarchical resource allocation

Effective link prediction in complex networks must balance predictive accuracy against the preservation of intrinsic topological integrity. Existing methods frequently introduce structural distortions by neglecting multi-scale organizational patterns. To address this limitation, we propose a structure-preserving link prediction method based on a Hierarchical Resource allocation framework (HR). This framework employs a three-layer architecture that integrates universal connectivity potentials, local similarity features, and mesoscale constraints. By incorporating k-shell gradients and community modularity, the framework modulates resource transmission efficiency to prioritize links aligned with the network’s hierarchical and modular organization. Comprehensive experiments across both real-world and synthetic networks demonstrate that the HR method achieves significant improvements in prediction accuracy compared to state-of-the-art baselines, while maintaining structural integrity across multiple organizational scales. These results validate that the framework enhances prediction accuracy without compromising the multi-scale structural fidelity of complex networks.