SCEC: A Novel Multi-Source Neighbor Aggregation Approach for Identifying Important Nodes in Complex Networks
Abstract
Identifying influential spreaders requires structural measures that capture local connectivityand a node’s position in the wider network. We propose structure and core enhanced centrality(SCEC), a ranking heuristic combining degree, clustering, coreness, and multi-orderneighbor information. The degree–clustering and first- and second-order degree–corenessterms form an SCC score, and SCEC aggregates the SCC values of immediate neighbors.We evaluate the method on eight empirical networks against archived sequential-updatesusceptible–infected–recovered (SIR) influence estimates at five transmission probabilitiesfrom 0.8 to 1.2 times a degree-moment threshold. Across the 40 network–transmissionsettings, SCEC had a mean Kendall correlation of 0.798 and a mean top-20% Jaccard similarityof 0.729, compared with 0.776 and 0.684 for NCC, the strongest principal baseline bythese averages. The performance varied by network and transmission setting. Ablationsindicate a larger contribution from final neighbor aggregation than from the clustering term,whose incremental benefit is limited in these data. Weight and selection-budget analysescharacterize the sensitivity within the tested settings.