Advancing resilience in infrastructure projects through machine learning-driven models

Traditional risk management enhances infrastructure utility but remains limited in addressing complexity and uncertainty. This has shifted attention towards resilience, particularly the readiness dimension, to improve early threat detection and prevention. Machine Learning (ML) offers opportunities to advance resilience modelling, yet empirically validated ML-enabled approaches, especially those using neural networks, are scarce, restricting accuracy, reliability, and applicability. This study develops a neural network-enabled resilience model optimized for training efficiency and predictive performance. By incorporating established feature importance techniques, the model improves accuracy, interpretability, and the identification of influential factors. The findings extend resilience typologies by ranking factor importance in critical infrastructure, highlighting ‘Operational resilience’ as the most significant determinant of project success. Practically, the model provides managers with clearer insights for decision-making, supporting earlier threat recognition and stronger disruption detection. The framework is adaptable across resilience contexts with appropriate industry-or platform-specific modifications.