Machine learning-based detection of workplace stress using wearable and multimodal data: a systematic literature review

Workplace stress is a significant concern, as it negatively impacts employee wellbeing and organizational productivity and is a major contributor to burnout and job turnover. Detecting stress in real-world work environments remains challenging; however, recent advances in machine learning and deep learning techniques offer promising solutions. Furthermore, the growing availability of multimodal data and wearable sensor technologies may facilitate individual stress tracking. In this paper, a systematic literature review is presented, focusing on machine learning and deep learning approaches for detecting workplace stress using wearable and multimodal data: (a) suggested machine learning and deep learning techniques are considered, (b) dataset characteristics and the sensor modalities employed are examined, (c) detection performance is reviewed, (d) workplace contexts and professional domains are discussed, and (e) gaps and future research directions are identified. The 20 selected studies show that machine learning and deep learning models applied to physiological, behavioral, and multimodal data can effectively detect workplace stress, particularly in high-risk occupations. However, limitations such as small sample sizes, limited dataset diversity, and minimal use of central nervous system signals (e.g., electroencephalogram) remain.