Predictive Analytics-Driven Decision Support System for Earned Value Management Using Ensemble Learning in Megaprojects

Abstract

This paper presents a predictive analytics-driven decision support system (DSS) for enhancing Earned Value Management (EVM) in large-scale infrastructure megaprojects. By integrating ensemble learning models such as Random Forests, Gradient Boosting Machines, and XGBoost, the proposed framework improves forecasting accuracy for cost variance (CV), schedule variance (SV), and Estimate at Completion (EAC). These models are trained on multivariate historical project data including baseline budgets, work performance indices, risk profiles, and change order logs. A key feature of this DSS is its capacity to dynamically recalibrate predictive models using real-time project control data, thereby enabling continuous optimization of decision-making under uncertainty. Comparative analysis against traditional parametric EVM forecasting techniques demonstrates significant improvements in Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE) across diverse megaproject scenarios. The system also incorporates explainable AI (XAI) techniques such as SHAP and LIME to enhance transparency in predictive reasoning, facilitating project stakeholder trust and actionable insights. The findings emphasize the strategic potential of ensemble machine learning in minimizing project overruns, improving planning reliability, and elevating the maturity of project analytics in complex, resource-intensive environments. This work offers a replicable blueprint for modernizing project control systems through data-driven intelligence.