Machine Learning-driven Strategies for Risk Interactions and Systemic Risk Management of Infrastructure Projects

Abstract

Infrastructure projects frequently fail to meet performance expectations, due to their inherent complexities, leading to delays, cost overruns, and safety concerns. Risk interactions and systemic risks are two key contributors to these challenges. Risk interactions occur when one risk amplifies the magnitude and/or the probability of another, such as extreme weather delaying work progress while also increasing safety incidents. Systemic risks arise from disruptions in one component that can lead to project-wide cascading disruptions, such as delays in excavation work impacting downstream work packages like backfilling and site grading. Previous studies investigated risk interactions and systemic risks separately, which often led to sup-optimal project performance. Additionally, existing models rely on complex simulations and rigid theoretical frameworks, limiting their practicality. In this respect, the research presented in this dissertation is aimed at developing machine learning (ML)- and optimization-based strategies to address both risk interactions and systemic risks in infrastructure projects. The proposed strategies enable practitioners to i) quantify and predict the combined impacts of risk interactions and systemic risks on the project performance, thereby improving the accuracy of risk assessment; and ii) implement adaptive solutions to rapidly restore key project performance targets. The findings of the current research highlight the value of integrating ML and optimization in decision-making, offering practical solutions to enhance project outcomes under the constraints of risk interactions and systemic risks. Importantly, the presented data-driven strategies are not meant to replace the existing project management tools in practice, but rather to complement them. Project managers should continue to exercise their professional judgment alongside these strategies to ensure efficient risk management. Overall, this work advances the understanding of risk management in large-scale infrastructure projects, providing data-driven approaches to improve project performance under complex risk conditions.