AI-Driven Project Portfolio and Business Analytics for Forecasting Cost Overruns, Schedule Delays, Resource Bottlenecks, and Enterprise Performance Across U.S. Industries

Abstract

Project portfolios are a primary mechanism through which U.S. organizations transform capital, technology, workforce capability, and public investment into productive capacity. Yet cost overruns, schedule delays, resource bottlenecks, and weak benefits realization can consume scarce resources and defer economically important outcomes. This study develops an explainable project portfolio and business-analytics framework for integrated early warning, severity forecasting, constrained resource allocation, and enterprise-performance assessment. The empirical design combines verified 2023 U.S. economic and public-project indicators with a reproducible synthetic panel of 18,000 projects across eight industries and ten project types. Models are trained on first- through third-quarter observations and evaluated on a strict fourth-quarter holdout. Logistic regression, random forest, and gradient boosting are compared for material cost overrun, schedule delay, resource bottleneck, and multidimensional project success; regression models estimate cost and delay severity and an enterprise-performance index. The framework adds probability calibration, global and local explanations, capacity-sensitive thresholding, portfolio optimization, stress testing, executive dashboards, and a traceable governance lifecycle. Results demonstrate that delivery risk is most informative when financial, scope, workforce, supplier, governance, and portfolio-congestion indicators are analyzed jointly. The proposed architecture does not replace project managers, engineers, executives, auditors, or public officials. It provides transparent evidence for earlier review, targeted intervention, and strategic allocation. The research offers a scalable foundation for strengthening U.S. enterprise productivity, capital efficiency, infrastructure and technology delivery, and organizational resilience.