ScholarWorks@성균관대학교

초록

Construction delays remain a critical challenge, often leading to cost overruns, contractual disputes, and project inefficiencies. This paper presents a proactive, data-driven framework designed to predict and manage delays specifically in highway earthwork projects, one of the most delay-sensitive and uncertainty-prone phases in infrastructure construction. Leveraging both structured and unstructured data, the proposed method combines pretrained language models (PLMs) and machine learning (ML) algorithms to accurately forecast delay types and durations. An automated probability-impact matrix (PIM) is incorporated to visualize and prioritize risks, enabling more strategic resource allocation. To ensure interpretability, the framework integrates SHapley Additive exPlanations (SHAP), offering stakeholders transparent insights into model predictions. Case studies demonstrate the framework's effectiveness in reducing manual workload and enhancing early decision-making in real-world construction settings. By embedding predictive intelligence into the delay management process, this approach aligns with the goals of Construction 4.0 and offers a scalable solution adaptable to broader industries. The proposed framework improves predictive accuracy and provides actionable insights that support more efficient and informed risk management.

키워드