Wearable Sensor-Based Fatigue Classification Under Diverse Thermal Conditions. Journal of Information Technology in Construction

Fatigue induced by physical exertion and environmental stress remains a critical safety concern in construction and other physically demanding industries. This paper investigates whether integrating wearable sensor data (EMG, HR, HRV) and thermal conditions (hot, room, cold) can improve real-time fatigue prediction. Physiological signals were collected using wearable sensors, processed through noise filtering and feature extraction, and classified using Random Forest Classifier and Extreme Gradient Boosting algorithms. The models demonstrated high predictive accuracy, achieving 80% for continuous fatigue levels and over 90% for categorical fatigue classes. These findings are particularly valuable for construction safety managers, occupational health researchers, and technology developers seeking proactive fatigue management solutions. Future research should focus on field validation of wearable systems and integration with site management platforms such as BIM for broader industry adoption.