Delineating infrastructure failure interdependencies and associated stakeholders through news mining: the case of Hong Kong’s water pipe bursts

Abstract

The failure of one infrastructure system could trigger cascading impacts on other interdependent infrastructures. In order to improve the management of diverse infrastructure systems, decision makers should be mindful of infrastructure failure interdependencies (IFIs) and associated stakeholders when the failure of a particular infrastructure occurs. Currently, approaches to identify IFIs and associated stakeholders rely heavily on expert knowledge or limited historical records. To complement the shortage of empirical evidence, a synthetic approach that exploits media news is proposed to delineate the patterns of IFIs and stakeholders associated with the initial infrastructure failure. The integrated approach collects and cleanses the corpus from news articles, prepares the domain knowledge components, recognizes the affected infrastructure and stakeholder entities, verifies the information captured, applies association rule learning to discover IFI chains, and adopts a network analysis to depict associated stakeholders. Incidents of bursting water pipes in Hong Kong are used as a case study to demonstrate the proposed approach with 2,828 news articles being collected and analyzed. Altogether, 18 one-order or second-order IFI rules are identified. Besides, 25 associated stakeholders are delineated from the news, and they are divided into three tiers according to their degree centralities. The findings provide insightful information to policymakers for helping to respond to the cascading effects among infrastructures and coordinate a wide spectrum of stakeholders who might be embroiled.