EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Hierarchical Bayesian Inference for Human Activity: Modeling Subject-Level Variability and Posterior Uncertainty

Uzma Yasmeen

Precision Journal of Applied Mathematics and Statistics · 2026

DOI: 10.67887/pd2601208003 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Precision Journal of Applied Mathematics and Statistics, 2026. Verfasst von Uzma Yasmeen. Die Arbeit nennt 76 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Introduction

The study proposed a Bayesian hierarchical model to examine longitudinal multilevel human activity data, based on the KU-HAR dataset of inertial sensor measurements of several subjects and activities. In contrast to classical methods of machine learning, where the accuracy of classification was the focus, the proposed model followed the objective to reproduce the underlying variance structure and to measure the uncertainty by full posterior inference. Methodology: A three-level hierarchy was used to divide variability into global, subject, and activity variability so that a stringent analysis of inter-individual and inter-class dynamics could be achieved. Markov Chain Monte Carlo (MCMC) sampling was used to estimate posterior distributions, and a convergence diagnostic was evaluated using trace evaluation, effective sample size, and Gelman-Rubin statistics.

Results

The findings indicate that the subject-level variability () was significantly higher than the activity-level variability (). It could be seen that the physiological and biomechanical differences among individuals were the ones that overshadowed the wearable sensor measurements. The posterior predictive tests showed that the generative fidelity was high, and replicated distributions were very close to the empirical ones.

Conclusion

The findings indicate the weakness of population-level models and the need to consider personalized probabilistic frameworks to have robust human activity recognition. The Bayesian method was not only able to capture latent uncertainty but also gave interpretable probabilistic parameters, which increased transparency and can be clinically used. The study builds upon the methodological basis of sensor-based behavioral modeling and contributes to the incorporation of hierarchical Bayesian inference into future healthcare monitoring and adaptive wearable systems.

Zitieren

APA

Yasmeen, U. (2026). Hierarchical Bayesian Inference for Human Activity: Modeling Subject-Level Variability and Posterior Uncertainty. Precision Journal of Applied Mathematics and Statistics. https://doi.org/10.67887/pd2601208003

Vancouver

Yasmeen U. Hierarchical Bayesian Inference for Human Activity: Modeling Subject-Level Variability and Posterior Uncertainty. Precision Journal of Applied Mathematics and Statistics. 2026. doi:10.67887/pd2601208003

BibTeX
@article{Yasmeen2026,
  author = {Yasmeen, Uzma},
  title = {Hierarchical Bayesian Inference for Human Activity: Modeling Subject-Level Variability and Posterior Uncertainty},
  journal = {Precision Journal of Applied Mathematics and Statistics},
  year = {2026},
  doi = {10.67887/pd2601208003}
}
Kostenlos starten →