Active Branch Reach Centrality: A Two-Hop Structural Measure for Influential Node Identification in Complex Networks
Abstract
Influential-node ranking often involves a trade-off between structural coverage and computational cost: global measures can be expensive, while very local scores may overlook useful spreading cues. We introduce Active Branch Reach Centrality (ABRC), a strict two-hop measure built from three quantities around a source node: degree, the number of unique distance-two nodes, and the number of first-hop neighbors that connect into that distance-two set. The final score weights the two-hop count by one half and the active-branch count by 7.5; the branch coefficient was selected once on USAir97 and then frozen. ABRC is evaluated against ten baseline centralities on seven real-world networks using a discrete susceptible-infected-recovered (SIR) process at five infection rates, with 500 Monte Carlo realizations per seed and rate. All baseline and component-ablation comparisons use the same cached SIR outcomes. Across the seven networks, ABRC obtains a mean Kendall $\tau _{b}$ of 0.7862 and an average rank of 3.857, while degree centrality obtains 0.7358. A four-component ablation shows that the active-branch term improves the full score over degree-plus-two-hop reach on four of six validation networks but reduces agreement on Dolphins and the US Power Grid; the validation-only mean branch increment is −0.0027 and the median is + 0.0133. The branch contribution is therefore useful but topology dependent rather than universally beneficial. SemiLocal and Katz retain higher overall mean correlations, and the sparse US Power Grid remains a clear limitation of strict two-hop locality.