AI Model Predicts Construction Worker Heat Stress with 95% Accuracy Using Wearable Data

A new study published on arXiv has unveiled deep learning models designed to predict heat stress among construction workers in Saudi Arabia, addressing an important safety issue. The research team monitored 19 workers using Garmin Vivosmart 5 smartwatches to gather data on their heart rate, heart rate variability, and oxygen levels. They found that an attention-based Long Short-Term Memory (LSTM) network performed better than the standard LSTM, achieving a remarkable testing accuracy of 95.40%, with fewer false positives and negatives. The model's precision, recall, and F1 scores were all around 0.982, making it suitable for integration into IoT safety systems and BIM dashboards, ultimately improving safety management in construction by translating real-time health data into actionable insights.