Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Ensemble Methods for Classification of Physical Activities from Wrist Accelerometry.

Alok Kumar Chowdhury1, Dian Tjondronegoro, Vinod Chandran

  • 11Science and Engineering Faculty, Queensland University of Technology, Brisbane, AUSTRALIA; and 2Institute of Health and Biomedical Innovation at QLD Centre for Children's Health Research, School of Exercise and Nutrition Sciences, Queensland University of Technology, Brisbane, AUSTRALIA.

Medicine and Science in Sports and Exercise
|April 19, 2017
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Enhancing Physically Active Leisure Participation for Children With Cerebral Palsy: A Randomized Controlled Trial.

Pediatrics·2026
Same author

Accelerometry-measured prolonged and interrupted sedentary behavior and cancer incidence and mortality: A cohort study of 91,292 UK Biobank participants.

PLoS medicine·2026
Same author

Compositional movement behaviours and preschool children's social-emotional development.

The international journal of behavioral nutrition and physical activity·2026
Same author

Environmental supportiveness, physical activity, and sedentary time in children with cerebral palsy.

Developmental medicine and child neurology·2026
Same author

Short and long-term effects of a play-based therapeutic exercise programme in children with non-cystic fibrosis bronchiectasis: A randomised controlled trial.

Journal of science and medicine in sport·2026
Same author

Evaluating the scale-up of the Play Active programme for children's physical activity in early childhood education and care services: a national type III hybrid effectiveness-implementation trial protocol.

BMJ open·2025

Ensemble learning algorithms significantly enhance physical activity recognition accuracy from accelerometer data compared to single classifiers. A custom ensemble model with weighted majority voting achieved the highest accuracy in most datasets.

Area of Science:

  • Wearable technology and sensor data analysis
  • Machine learning applications in healthcare
  • Human activity recognition

Background:

  • Accurate physical activity recognition is crucial for health monitoring and disease management.
  • Single machine learning classifiers often face limitations in recognizing complex human activities.
  • Ensemble learning offers a promising approach to improve classification performance.

Purpose of the Study:

  • To evaluate if ensemble learning algorithms improve physical activity recognition accuracy over single classifiers.
  • To compare the performance of conventional ensemble methods (bagging, boosting, random forest) against a custom ensemble model.
  • To assess the efficacy of different decision fusion techniques within the custom ensemble.

Main Methods:

Related Experiment Videos

  • Utilized three independent datasets of wrist-worn accelerometer data.
  • Implemented a four-step classification framework: preprocessing, feature extraction, normalization, and classifier training/testing.
  • Employed leave-one-subject-out cross-validation and compared classifiers based on average F1 scores.

    • Main Results:

    • Ensemble learning methods consistently outperformed individual classifiers across all datasets.
    • Random forest models demonstrated high and consistent activity recognition accuracy.
    • The custom ensemble model with weighted majority voting achieved the highest classification accuracy in two of the three datasets.

    Conclusions:

    • Combining multiple classifiers via ensemble learning significantly boosts activity recognition accuracy.
    • Ensemble methods, particularly custom models with effective fusion strategies, are superior to single classifiers for accelerometer-based activity recognition.