On Selecting the Relevant Metrics of Network Robustness

Abstract

This paper deals with selecting the most relevant metrics with which to measure the robustness of a network. Although earlier efforts have also attempted to do this, there is still no consensus on how to define a single robustness metric. Instead, a large set of metrics regarding the structural, fragmentation, connectivity and centrality properties of a graph have been used. Here, we propose a novel methodology based on the Principal Component Analysis to calculate a single value Robustness* (R*). This is also a consistent way of analyzing how a network behaves under a severe removal of elements. Results show how to select the most relevant metrics for robustness and how to apply them in heterogeneous topologies.