Building and Environment, 289, 114041
Health|구충완
Climate-change adaptation to extreme heat on construction sites: A wearable-free, simplified machine-learning model for predicting workers’ heat strain
2026International Journals
Research AreaHealth
Professor구충완
AuthorsLee, J., Seo, S., Choi, D., Choi, Y., and Koo, C. (Corresponding Author)
Publication Year2026
Journal (Volume, Issue, Pages)Building and Environment, 289, 114041
AbstractConstruction workers are increasingly vulnerable to heat strain due to prolonged outdoor exposure and the intensifying effects of global warming. To support climate-adaptation strategies at jobsites, this study developed a wearable-free, simplified machine-learning (ML) model capable of delivering real-time, individualized heat strain forecasts. The model overcomes the acceptance barriers associated with heart-rate monitoring devices by relying exclusively on readily obtainable environmental, personal, and work-related variables. A living-lab study was conducted at two construction sites in South Korea, collecting 95,340 records from 67 workers. Even under similar climate conditions, heat strain varied significantly among individuals, reflecting the combined influence of personal attributes and occupational demands. The proposed model demonstrated high predictive accuracy (MAE = 0.0899 °C), while Shapley Additive Explanations (SHAP) analysis identified cumulative work time, occupation, work period, apparent temperature, and age as the most influential predictors. The model also exhibited strong classification performance (Precision = 0.82, Recall = 0.77, F1-score = 0.79), confirming its reliability for threshold-based risk prediction. Building on these findings, a process-based system prototype was designed to operationalize the model’s outputs through an integrated workflow encompassing data collection, real-time inference, alert notifications, and periodic model re-training—all without requiring body-worn sensors. This prototype enables proactive, worker-tailored heat strain management that is both acceptable to workers and practically deployable on construction sites. The proposed approach establishes a scalable, evidence-based foundation for heat-safety protocols and offers a practical template for climate-change adaptation in labor-intensive industries.
Construction WorkerIndividual Heat StrainPredicted Heat Strain (PHS)Machine-Learning AlgorithmShapley Additive Explanation (SHAP)Process-Based Prototype