Systematic Review of Explainable Artificial Intelligence (XAI) Applications in Predicting and Controlling Delay and Cost Overrun in Construction Projects: Methods, Challenges, and a Proposed Decision-Making Framework

Systematic Review of Explainable Artificial Intelligence (XAI) Applications in Predicting and Controlling Delay and Cost Overrun in Construction Projects: Methods, Challenges, and a Proposed Decision-Making Framework

Shahabeddin Khoshnoudi1

1) M.Sc. Student, Department of Civil Engineering, Construction Management Orientation, Rahman Institute of Higher Education, Ramsar,

IOI Code : ABKD-ABKEBD
Publication : Fourth International Conference on applied researches in civil engineering, architecture and urban planning (icarcau.com)
Abstract :
    Time delay and cost overrun are among the most persistent issues in construction project management. The complexity of interrelationships among project characteristics, market conditions, contractor performance, contractual changes, and execution risks has limited the effectiveness of linear methods and purely experience-based judgments. In recent years, machine learning models have improved prediction accuracy; however, the black-box nature of many such models reduces managers trust, the defensibility of decisions, and the feasibility of translating predictions into corrective actions. This study aims to systematically review the literature on explainable artificial intelligence (XAI) in predicting and controlling delay and cost overrun, and to propose a decision-support framework applicable to construction management. The results indicate that tree-based ensemble models, particularly gradient boosting, random forest, and hybrid models, are the most widely used, with SHAP being the most common explainability tool. Despite advances in identifying cost and time drivers, most research remains focused on prediction, while uncertainty quantification, counterfactual explanations, cross-project validation, and direct linkage to managerial actions are less developed. Accordingly, a six-layer framework comprising data governance, probabilistic modeling, multi-level explainability, action prioritization, human oversight, and feedback learning is proposed.

Keywords : construction project management project delay cost overrun explainable artificial intelligence (xai) machine learning risk management