A Framework for Leveraging Real-Time Predictive Risk Analytics in Large-Scale Infrastructure Projects to Mitigate Cost Overruns and Schedule Delays

• Abbas Ahsun

  Texas University

  Author

Large-scale infrastructure projects continue to experience persistent cost overruns and schedule delays despite decades of risk management research and practice. While traditional estimation approaches rely on static historical averages and subjective expert judgment, they fundamentally fail to capture the nonlinear, interdependent nature of risks inherent in complex project environments. This study addresses this gap by developing and validating a machine learning-based predictive risk analytics framework that integrates LightGBM, explainable artificial intelligence (XAI), and digital twin simulation to enable real-time cost and schedule risk forecasting. Using a comprehensive dataset of 1,271 change order records from transportation infrastructure projects spanning 2010–2025, the proposed framework achieved a prediction accuracy of 89.4% in classifying change order impact severity, with a mean absolute percentage error of 6.2% for cost overrun magnitude prediction. The LightGBM-based model demonstrated a 23% improvement in predictive accuracy over conventional static estimation methods, while SHAP-based feature importance analysis identified material price volatility, scope-of-work changes, and construction progress at change order issuance as the top three risk predictors. The framework provides project managers with actionable risk alerts with a lead time of approximately 12-15 days, enabling proactive contingency allocation and schedule adjustments. This research contributes a replicable, explainable, and practically deployable predictive risk analytics framework that bridges the gap between theoretical risk modeling and operational decision-making in infrastructure project delivery.

• PDF

Copyright (c) 2026 Abbas Ahsun (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.