Centralidade de proximidade por múltiplos caminhos disjuntos

Abstract

Traditional centrality metrics consider only shortest paths, neglecting the existence of bit-longer paths between nodes in a network, which can be strategic to maintain network connectivity. This work proposes the disjoint multipath closeness centrality, which extrapolates the traditional closeness by considering multiple shortest and quasi-shortest disjoint paths. The conectivity factor ϕ limits the desirable number of multiple disjoint paths. The idea is to identify nodes that are simultaneously multiply-connected and close to all the remaining nodes. Such nodes are important do perform tasks that require higher availability. We evaluate the proposed metric through comparisons with other centrality metrics. Our results confirm that the most central nodes according to the proposed metric are more accessible when failures happen in random nodes in the network. Moreover, our results show that the use of multiple disjoint paths can reclassify at least 59% of nodes in the evaluated datasets. Hence, our proposed metric is able to identify better-connected nodes and assign them more importance.