CITY NETWORK RESILIENCE QUANTIFICATION UNDER SYSTEMIC RISKS: A HYBRID MACHINE LEARNING-GENETIC ALGORITHM APPROACH

Abstract

Disruptions due to either natural or anthropogenic hazards significantly impact the operation of critical infrastructure networks because they may instigate network-level cascade (i.e., systemic) risks. Therefore, quantifying and enhancing the resilience of such complex dynamically evolving networks ensure minimizing the possibility and consequences of systemic risks. Focusing only on robustness, as one of the key resilience attributes, and on transportation networks, key critical infrastructure, the current study develops a hybrid complex network theoretic-genetic algorithms analysis approach. To demonstrate the developed approach, the robustness of a city transportation network is quantified by integrating complex network theoretic topology measures with a dynamic flow redistribution model. The network robustness is subsequently investigated under different operational measures and the corresponding absorptive capacity thresholds are quantified. Finally, the robustness of the network under different failure scenarios is evaluated using genetic algorithms coupled with k-means clustering to classify the different network components. The hybrid approach developed in the current study is expected to facilitate optimizing potential systemic risk mitigation strategies for critical infrastructure networks under disruptive events.