Federated learning for nurse stress prediction using wearable sensors: Integrating biomechanical data

In today’s fast-paced work environments, accurately predicting stress levels is essential for effective healthcare workforce management, particularly among nurses in high-pressure settings. Despite the availability of various mental health initiatives, timely detection of stress remains challenging due to concerns over sensitive personal data privacy. To address this, we propose a federated learning (FL) framework that utilizes artificial intelligence (AI) to predict nurse stress levels by integrating distributed biomechanical data from wearable sensors, thereby preventing data leakage. Biometric features from datasets at each FL client are extracted and used to train local neural network (NN) models. After several aggregation rounds, the global model converges to predict nurse stress levels. Simulations demonstrate the effectiveness of our method, achieving over 90% prediction accuracy, which enhances the feasibility of privacy-preserving stress monitoring and offers scalable solutions for occupational health management.