A Hybrid AI Framework for Time & Cost Performance Forecasting in Oil and Gas Project Management

Abstract

Oil and gas (O&G) projects frequently experience significant time delays and cost overruns due to uncertainty, project complexity, and volatile market and regulatory conditions. Improving the reliability of early performance forecasts is therefore critical for effective project planning and risk mitigation. This paper introduces a hybrid artificial intelligence decision-support framework that combines Bayesian Networks, Extreme Gradient Boosting, and Simulated Annealing (BN–XGBoost–SA) to predict time and cost overruns in O&G projects. The framework integrates expert knowledge and historical data to model project risks, capture complex nonlinear patterns, and optimize predictive performance, while embedding physics-based models to ensure operational realism. A systematic implementation process is developed, covering data preparation, model development, validation with explainable AI, and real-time deployment through interactive dashboards. The proposed approach enables project managers to identify key drivers of overruns, quantify uncertainty, and evaluate the impact of mitigation actions before critical decisions are made. Case-based simulations demonstrate predictive accuracy of up to 92% and indicate potential cost savings of 20–30% through proactive intervention. The framework provides managers with an interpretable, adaptive, and practical tool for improving schedule and cost control in large-scale O&G projects.