Longitudinal estimation of stress-related states through bio-sensor data

Purpose The authors aim to develop a conceptual framework for longitudinal estimation of stress-related states in the wild (IW), based on machine learning (ML) algorithms that use physiological and non-physiological bio-sensor data. Design/methodology/approach propose consisting four blocks: (1) identification; (2) validation; (3) measurement (4) visualization. implement each step proposed framework, using example Gaussian mixture model (GMM) K-means algorithm. These ML are trained data 18 workers from public administration sector who wore biometric devices about two months. Findings confirm convergent validity IW. Empirical analysis suggests two-cluster models achieve five-fold cross-validation accuracy exceeding 70% identifying stress. Coefficient decreases three-cluster achieving around 45%. conclude identification may serve derive measures. Research limitations/implications Proposed guide researchers creating validated indicators. At same time, sensing stress through is limited because subject-specific reactions stressors. Practical implications Longitudinal indicators allow long-term impact coming external environment states. Such can become an integral part mobile/web/computer applications supporting management programs. Social Timely excessive improve individual well-being prevent development diseases. Originality/value study develops novel data, given scientific knowledge emergent state.