A machine learning-based forecasting model for personal maximum allowable exposure time under extremely hot environments

Abstract

• Continuous work time control is crucial for preventing heat-related illnesses outdoors. • Machine learning was applied to forecast personal maximum allowable exposure time. • The proposed model improved forecast accuracy by incorporating biometric characteristics. • The most reliable model achieved a low mean absolute error (MAE) of just 11 seconds. • Personalized heat stress management strategy is now feasible for construction workers. As global warming leads to an increase in the frequency and intensity of heatwaves, protecting outdoor workers from heat-related illnesses becomes paramount. To address this challenge, this study aimed to forecast personal maximum allowable exposure time by considering individual differences under extremely hot environments. To enhance the prediction accuracy of the proposed approach, a machine learning-based error correction model was developed in conjunction with the PHS_HR method (i.e., predicted heat strain index using real-time heart rate), reflecting dynamic changes in personal biometric characteristics (e.g., heart rate, body fat percentage, etc.) as well as environmental conditions and exposure time. Among the developed models, the multi-layer perceptron (MLP) algorithm demonstrated a high degree of reliability in forecasting personal maximum allowable exposure time, achieving a mean absolute error (MAE) of 0.19 minutes (11 seconds), compared to the existing PHS_HR method (MAE: 5.05 minutes). This study highlights the potential benefits of data-driven computational methods in occupational health and safety management. By providing more accurate predictions of personal maximum allowable exposure time under extremely hot environments, the risk of heat-related illnesses outdoors can be effectively mitigated. As heatwaves become more prevalent, the proposed approach offers a more valuable tool to ensure safer work environments.