Robustness in clustering-based weighted inter-connected networks

We study the robustness of symmetrically coupled and clustering-based weighted heterogeneous inter-connected networks with respect to load-failure-induced cascades. This is done under the assumption that the flow dynamics are governed by global redistribution of loads based on weighted betweenness centrality. Our results indicate that no weighting bias should be assigned to inter-links when calculating shortest path between node pairs under the clustering-based weighting scheme; i.e., inter-links shall be treated no differently than intra-links. In contrast with local load redistribution cases, we show that increasing connectivity is preferred for the robustness against global load redistribution-based cascading failures in clustering-based weighted inter-connected networks. Furthermore, comparisons among weighting schemes reveal that, both the clustering-based and degree-based schemes outperform the random one in the sense of requiring lower initial and total investments required to ensure robustness. We also find that clustering-based scheme outperforms degree-based one in terms of requiring lower initial investments. Except in a limited range where weighting is heavily suppressed, clustering-based scheme is shown to outperform degree-based one in terms of total investments. Finally, when there exists a hard investment budget constraint, clustering-based weighting scheme would be a better choice against a two-nodes-induced failure than the degree-based weighting, and the clustering-based scheme is more stable than degree-based scheme against one-or-two-nodes-induced failure. We expect our findings to be significantly useful in designing real-world weighted inter-connected networks that are robust against load-failure-induced cascades.