Identifying influential nodes through hierarchical k-shell and extended neighborhood integration

Overview

This study addresses the node influence identification problem in complex networks through a hybrid algorithmic approach. The research targets limitations in existing methods concerning the trade-off between computational efficiency and ranking accuracy. The proposed HKEN algorithm combines hierarchical k-shell decomposition with extended neighborhood analysis to integrate global network structure with local topological properties.

Methods and approach

The HKEN algorithm operates through three primary mechanisms. First, hierarchical k-shell decomposition is refined through dynamic weight computation that incorporates both degree and k-shell values to optimize the coarse-grained structure of the base k-shell method. Second, the node neighborhood scope is expanded beyond immediate neighbors, with a local clustering coefficient threshold establishing constraints on information transmission distance. Third, an influence aggregation strategy employs Jaccard similarity coefficient calculations to synthesize influence scores. The methodology integrates global structural features derived from k-shell analysis with local topological information from extended neighborhoods.

Results

Experimental validation was conducted across 10 real-world networks and compared against 12 benchmark methods. The HKEN algorithm demonstrated improved alignment with Susceptible-Infected-Recovered (SIR) epidemic propagation models relative to comparison methods. Top-ranked nodes identified by HKEN exhibited enhanced propagation capability in network diffusion processes. Performance metrics confirm superior identification accuracy when validated against simulated cascade dynamics.

Implications

The results indicate that integrating global decomposition methods with local neighborhood information yields more accurate influence rankings than methods relying solely on individual approaches. The improved performance suggests that node influence operates through both macro-level network positioning and micro-level local interactions, requiring hybrid analysis frameworks. The computational efficiency gains from optimized k-shell processing alongside localized metrics address practical constraints in large-scale network analysis.