Related Experiment Video
Updated: Feb 28, 2026
11:21
Methodology for Establishing a Community-Wide Life Laboratory for Capturing Unobtrusive and Continuous Remote Activity and Health Data
Published on: July 27, 2018
8.9K
Unveiling the Digital Phenotype of Physical Activity Behavior in Community-Dwelling Older Adults Using Machine
Anas Abdulghani1, Kim Daniels2,3, Bruno Bonnechère1,2,3
1Technology-Supported and Data-Driven Rehabiltitation, Data Sciences Institute, Hasselt University, 3590 Diepenbeek, Belgium.
Bioengineering (Basel, Switzerland)
|February 27, 2026
Summary
Machine learning accurately predicts physical activity (PA) in older adults. Self-efficacy and depression scores are key cross-sectional predictors, while past step data best forecasts future PA longitudinally.
Area of Science:
- Gerontology
- Health Informatics
- Data Science
Background:
- Physical activity (PA) is crucial for older adults' health and well-being.
- Predicting and understanding PA patterns in community-dwelling older adults is essential for targeted interventions.
Purpose of the Study:
- To apply machine learning (ML) to predict PA patterns in older adults.
- To identify key factors influencing PA behaviors using cross-sectional and longitudinal data.
Main Methods:
- Cross-sectional analysis utilized Linear Regression, Logistic Regression, Elastic Net, and Light Gradient Boosting Machine (LightGBM).
- Longitudinal analysis employed LightGBM, Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM) on 14-day data.
- Wearable sensor data (step count) and validated questionnaires (ESES, GDS, IPAQ) were used.
Main Results:
- Cross-sectional analysis identified Exercise Self-efficacy Scale (ESES) and Geriatric Depression Scale (GDS) as significant predictors.
- Longitudinal analysis showed a seven-day step count sequence yielded the best daily PA forecast.
- LightGBM and LSTM models demonstrated strong predictive capabilities.
Conclusions:
- Combining wearable sensor data with ML/deep learning offers valuable insights into older adults' PA.
- Psychological factors (self-efficacy) and past activity data are critical for predicting future physical activity levels.
