Related Experiment Video
Updated: Nov 2, 2025
08:05
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
10.9K
Personalised Accelerometer Cut-point Prediction for Older Adults' Movement Behaviours using a Machine Learning
Nonso Nnamoko1, Luis Adrián Cabrera-Diego2, Daniel Campbell3
1Department of Computer Science, Edge Hill University, Ormskirk, L39 4QP, United Kingdom.
Computer Methods and Programs in Biomedicine
|June 12, 2021
Summary
This study introduces a machine learning method to personalize physical activity intensity cut-points for older adults, improving accuracy over generic methods. The personalized approach consistently outperformed the state-of-the-art in predicting activity intensity without prior data.
Area of Science:
- Gerontology
- Biomedical Engineering
- Data Science
Background:
- Objective physical activity assessment in older adults commonly uses body-worn accelerometers.
- Generic accelerometer cut-points for activity intensity lack individual specificity, leading to varied results.
- Current methods assume a 'one size fits all' approach, which is inadequate for diverse older adult populations.
Purpose of the Study:
- To propose and evaluate a machine learning method for personalizing accelerometer-based physical activity intensity cut-points in older adults.
- To address the limitations of generic cut-points by developing individualized activity intensity thresholds.
- To improve the accuracy and reliability of objective physical activity measurement in older adults.
Main Methods:
- Collected accelerometry data from 33 older adults using GENEActive (wrist) and ActiGraph (hip) devices.
- Applied ROC analysis to generate personalized cut-points, optimizing for sedentary behavior and moderate-to-vigorous physical activity.
- Utilized an additive regression algorithm trained on participant biodata to predict cut-points, evaluating with Mean Absolute Error, Root Mean Square Error, and Standard Error of Estimation.
Main Results:
- The personalized machine learning approach demonstrated consistent superiority over the state-of-the-art across all four cut-points and both accelerometer devices.
- Significant reductions in Standard Error of Estimation were observed for both ActiGraph and GENEActive devices compared to existing methods.
- The proposed method achieved better prediction accuracy and goodness of fit, indicating enhanced reliability.
Conclusions:
- Personalized activity intensity cut-points can be accurately predicted using machine learning without requiring prior accelerometry data.
- The findings are highly promising for objective physical activity assessment in older adults, offering a more individualized approach.
- Further research with expanded datasets is recommended to broaden the applicability of this personalized method.
