A Data-Driven Framework for Engineering Project Risk Assessment Under Schedule and Cost Uncertainty

Abstract

Engineering projects are frequently exposed to interconnected schedule and cost uncertainties that can reduce the reliability of deterministic planning, contingency allocation, and project-control decisions. This study developed and quantitatively evaluated a data-driven framework for engineering project risk assessment by integrating historical project performance, statistical modeling, schedule-cost interdependence, Monte Carlo simulation, sensitivity analysis, probabilistic contingency estimation, and predictive validation. The analytical sample consisted of 186 completed engineering projects containing 7,842 scheduled activities and 3,416 documented risk observations across infrastructure, industrial, energy, transportation, and utility environments. Mean actual project duration was 29.6 months compared with a planned duration of 24.8 months, representing an average schedule overrun of 19.4%, while mean final cost increased from $47.6 million to $54.9 million, corresponding to a 15.3% cost overrun. Only 33.3% of projects were completed within the original schedule and 42.5% remained within the approved budget. Monte Carlo simulation produced P80 targets of 34.7 months and $63.8 million, indicating substantially greater schedule and financial requirements than deterministic baselines. Schedule and cost overruns were strongly correlated (r = 0.74), and the probability of simultaneous schedule and cost overrun reached 49.1%, whereas the probability of achieving both baseline targets was only 23.5%. Procurement delay was the strongest independent schedule predictor (β = 0.31), while scope change was the strongest cost predictor (β = 0.29). High-complexity projects recorded 29.7% schedule overrun, 23.8% cost overrun, 96.4% contingency utilization, and 68.5% joint overrun probability. Predictive validation demonstrated strong model performance, with R² values of 0.84 for schedule and 0.82 for cost, MAPE values of 7.8% and 8.3%, and out-of-sample accuracy of 89.2% and 86.5%, respectively. The findings demonstrate that integrating probabilistic uncertainty, risk dependence, sensitivity, contingency adequacy, and predictive validation provides a more comprehensive quantitative representation of engineering project schedule and cost exposure than deterministic assessment alone.