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 Concept Videos

Survival Tree01:19

Survival Tree

Survival trees are a non-parametric method used in survival analysis to model the relationship between a set of covariates and the time until an event of interest occurs, often referred to as the "time-to-event" or "survival time." This method is particularly useful when dealing with censored data, where the event has not occurred for some individuals by the end of the study period, or when the exact time of the event is unknown.
 Building a Survival Tree
Constructing a survival tree begins...

You might also read

Related Articles

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

Sort by
Same author

Twelve-year trends in sex and age of patients with COPD and chronic respiratory failure undergoing pulmonary rehabilitation.

European journal of internal medicine·2026
Same author

Why almost all ML models for medicine are wrong-and what we need for evidence-based medical AI.

International journal of medical informatics·2026
Same author

A multimodal EHR-based phenotyping framework integrating consensus clustering and transformer-based clinical NLP: application to autoimmune gastritis.

International journal of medical informatics·2026
Same author

Evaluating large language models for structuring cardiology reports: a real-world clinical study on patient subtyping and trial recruitment.

International journal of medical informatics·2026
Same author

ViTMARE - A Vision Transformer Pipeline for Anomaly Detection in 3D Brain MRI.

Studies in health technology and informatics·2026
Same author

Environmental Personal Exposure Clusters to Investigate Multiple Sclerosis and Amyotrophic Lateral Sclerosis Progression.

Studies in health technology and informatics·2026
Same journal

Informed consent and ethical considerations in AI for dentistry and medicine: a scoping review.

International journal of medical informatics·2026
Same journal

An explainable machine learning framework for computable physiologic risk representation in preanesthetic assessment: Development and external validation.

International journal of medical informatics·2026
Same journal

Large language models for automated PRISMA 2020 adherence checking.

International journal of medical informatics·2026
Same journal

Artificial intelligence-enhanced nurse navigation for monitoring and care of long COVID.

International journal of medical informatics·2026
Same journal

Chain-of-verification prompting for NIH stroke scale extraction using small and frontier large language models.

International journal of medical informatics·2026
Same journal

Pilot evaluation of smart glasses for communication workflow and task performance in prehospital care.

International journal of medical informatics·2026
See all related articles

Related Experiment Video

Updated: May 9, 2026

Using a Split-belt Treadmill to Evaluate Generalization of Human Locomotor Adaptation
08:04

Using a Split-belt Treadmill to Evaluate Generalization of Human Locomotor Adaptation

Published on: August 23, 2017

How improper dataset split hinders model generalizability: a systematic comparison in Human activity recognition and

Samuele Pe1, Arianna Dagliati1, Enea Parimbelli2

  • 1Department of Electrical, Computer, and Biomedical Engineering, University of Pavia 27100 Pavia, Italy.

International Journal of Medical Informatics
|May 7, 2026
PubMed
Summary
This summary is machine-generated.

Improper data splitting in Human Activity Recognition (HAR) models inflates performance estimates. Using cross-subject splits is crucial for reliable generalization in healthcare AI.

Keywords:
Bias-variance tradeoffCross-subjectDeep learningKinect cameraMachine learning

More Related Videos

Assessing the Accuracy of Fitness Smartwatch Data for Cardiovascular and Physical Activity Monitoring: A Validation Study in Digital Health
05:51

Assessing the Accuracy of Fitness Smartwatch Data for Cardiovascular and Physical Activity Monitoring: A Validation Study in Digital Health

Published on: February 21, 2025

Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
06:49

Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment

Published on: December 11, 2015

Related Experiment Videos

Last Updated: May 9, 2026

Using a Split-belt Treadmill to Evaluate Generalization of Human Locomotor Adaptation
08:04

Using a Split-belt Treadmill to Evaluate Generalization of Human Locomotor Adaptation

Published on: August 23, 2017

Assessing the Accuracy of Fitness Smartwatch Data for Cardiovascular and Physical Activity Monitoring: A Validation Study in Digital Health
05:51

Assessing the Accuracy of Fitness Smartwatch Data for Cardiovascular and Physical Activity Monitoring: A Validation Study in Digital Health

Published on: February 21, 2025

Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
06:49

Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment

Published on: December 11, 2015

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Computer Vision
  • Healthcare Technology

Background:

  • Human Activity Recognition (HAR) and exercise assessment models are vital for healthcare applications like clinical evaluation and remote monitoring.
  • Real-world applicability hinges on generalizing to diverse subjects, yet many studies use non-cross-subject (NCS) splits, inflating performance estimates.
  • This practice risks misleading clinical trust due to inter-individual variability in movement patterns.

Purpose of the Study:

  • Investigate the impact of non-cross-subject (NCS) versus cross-subject (CS) data splits on machine learning and deep learning model performance.
  • Analyze how data splitting strategies and training-test set differences influence predictive variance and model stability.
  • Assess performance across tasks of varying complexity in Human Activity Recognition.

Main Methods:

  • Experiments utilized the large-scale NTU RGB+D 120 and IntelliRehabDS datasets for HAR and rehabilitation tasks.
  • Evaluated 12 machine learning and deep learning models using a simulation-based approach to compare NCS and CS split performance.
  • Employed predictive variance decomposition via Generalized Linear Mixed-Effects models to link split strategy to model stability.

Main Results:

  • Non-cross-subject (NCS) splits consistently overestimated model performance, especially with increasing task and model complexity.
  • Deep learning architectures showed significantly higher NCS performance compared to CS splits.
  • Greater subject differences between training and test sets increased predictive instability, while CS splits promoted more generalizable representations.

Conclusions:

  • Incorrect dataset splits can exaggerate generalization capabilities and undermine trust in AI models for rehabilitation and healthcare.
  • This study offers empirical evidence and methodological guidance for robust evaluation of computer vision-based rehabilitation models.
  • Promoting reproducible and trustworthy AI deployment is essential for broader healthcare applications.