Machine learning and optimization strategies for infrastructure projects risk management

Infrastructure projects often encounter significant performance challenges due to their inherent complexities. Two primary factors contributing to these challenges are risk interactions—occur when one risk amplifies another—and systemic risks, where disruptions in individual components can cascade into project-wide failures. Despite their critical importance, the combined impacts of these risks remain underexplored, particularly through practical and scalable methodologies. This study introduces an integrated machine learning (ML) and optimization-driven approach for assessing and mitigating these combined impacts on infrastructure project performance. Historical project data is leveraged to predict the performance impacts, measured through key performance indicators (KPIs). To enhance predictive accuracy and minimize computational costs, genetic algorithm-based hyperparameter tuning is employed, outperforming traditional methods such as grid search. Building on these predictions, multi-objective optimization is applied to devise effective response strategies that improve the project KPIs while adhering to predefined constraints. The utility of the proposed approach is illustrated through a demonstration application, showcasing its ability to generate optimized schedules and risk registers. These outputs offer actionable insights and decision support tools for project managers. The study contributes a scalable and practical solution that enhances the performance of infrastructure projects under the combined impacts of risk interactions and systemic risks.