Efficient labor forecasting remains a critical yet underexplored challenge in industrialized construction, particularly in Pre-Engineered Building (PEB) fabrication environments characterized by repetitive workflows, tradespecific labor dynamics, and cost-sensitive schedules. This study proposes a classification-enhanced machine learning framework that integrates Random Forest regression with real-time biometric labor data to predict workforce requirements. Projects are systematically classified using fixed thresholds for labor intensity and variability, yielding behaviorally distinct groups that improve model specialization and reduce forecast variance. Dedicated Random Forest models are trained for each classification group, leveraging structured biometric attendance logs to ensure input data fidelity. Model performance is assessed using RMSE and MSE metrics, while a profitability-based evaluation quantifies financial outcomes associated with prediction deviations. Experimental results show that over 88% of forecasts fall within an acceptable
Key Words
biometric data; classification; labor forecasting; machine learning in construction; Pre-Engineered Building (PEB); profitability analysis; random forest
Address
Ringle Raja, Hemalatha — Department of Civil Engineering, Karunya Institute of Technology and Sciences, Coimbatore, India
Elizabeth Amudhini Stephen — Department of Mathematics, Karunya Institute of Technology and Sciences, Coimbatore, India
Athish — Department of AI and ML Karunya Institute of Technology and Sciences, Coimbatore, India
Charles Climent Fliex — Elshaddai Engineering Private Limited, Chennai, India
PDF Viewer
Preview uses the same access rules as Full Text PDF (subscription, purchase, or open access).
