Disjoint multipath closeness centrality.

Traditional centrality metrics consider only shortest paths, neglecting alternative paths that can be strategic to maintain network connectivity. This paper proposes the disjoint multipath closeness centrality, a new metric to compute node centrality that extrapolates the traditional closeness to consider multiple shortest and disjoint quasi-shortest paths. The idea is to identify nodes that are close to all other nodes and are multiply-connected, which is important to perform high availability tasks. We limit the number of multiple disjoint paths using a connectivity factor φ . We comparatively investigate the correlation between our metric, the traditional closeness, and the information centrality using social and communication networks. We also assess the node ranking obtained by each metric and evaluate node reachability when one or multiple network failures occur. The results show that our metric maintains high concordance with the other closeness metrics but it can reclassify at least 59% of nodes in the evaluated networks. Our metric indeed identifies better-connected nodes, which remain more accessible when failures happen. [ABSTRACT FROM AUTHOR]