Tackling inter-subject variability in smartwatch data using factorization models

• Published in: Scientific reports Vol. 15; no. 1; pp. 26704 - 12
• Main Authors: Naseri, Arman, Tax, David M. J., van der Bilt, Ivo, Reinders, Marcel
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 23.07.2025; Nature Publishing Group; Nature Portfolio
• Subjects: 639/705/117; 639/705/531; Adult; Algorithms; Cardiovascular disease; Classification; Data acquisition; Exercise; Female; Heart rate; Heart Rate - physiology; Humanities and Social Sciences; Humans; Inter-subject variability; Light; Machine learning; Male; Monitoring, Physiologic - methods; multidisciplinary; Neural networks; Oxygen Saturation; Personal health; Physiology; REM sleep; Science; Science (multidisciplinary); Sleep - physiology; Smartwatch; Smartwatches; Standard scores; Time series; Wearable computers; Wearable Electronic Devices; Neural networks; Inter-subject variability; Smartwatch; Machine learning
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-025-12102-7

Smartwatches enable longitudinal and continuous data acquisition. This has the potential to remotely monitor (changes) of the health of users. However, differences among subjects (inter-subject variability) limit a model to generalize to unseen subjects. This study focused on binary classification tasks using heart rate and step counter from smartwatches, including night/day and inactive/active classification, as well as sleep and SpO2-related (oxygen saturation) tasks. To address inter-subject variability, we explored different transforming and normalization regimes for time series including per-subject and population-based strategies. We propose a modified factorized autoencoder, which separates the data into two latent spaces capturing class-specific and subject-specific information. Our proposed generalized factorized autoencoder and triplet factorized autoencoder improved classification accuracy over the baseline from 74.8 (± 10.5) to 83.1 (± 5.1) and 83.4 (± 5.3), respectively, for night/day classification, gains for inactive/active classification were modest, improving from 84.3 (± 9.4) to 86.9 (± 4.4) and 86.6 (± 4.3), respectively. Our study highlights challenges of handling inter-subject variability in smartwatch data and how factorization models can be used to enable more robust and personalized health monitoring solutions for diverse populations.