We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Quantitative Biology > Quantitative Methods

Title: Using unsupervised machine learning to quantify physical activity from accelerometry in a diverse and rapidly changing population

Abstract: Accelerometers are widely used to measure physical activity behaviour, including in children. The traditional method for processing acceleration data uses cut points to define physical activity intensity, relying on calibration studies that relate the magnitude of acceleration to energy expenditure. However, these relationships do not generalise across diverse populations and hence they must be parametrised for each subpopulation (e.g., age groups) which is costly and makes studies across diverse populations and over time difficult. A data driven approach that allows physical activity intensity states to emerge from the data, without relying on parameters derived from external populations, and offers a new perspective on this problem and potentially improved results. We applied an unsupervised machine learning approach, namely a hidden semi Markov model, to segment and cluster the accelerometer data recorded from 279 children (9 to 38 months old) with a diverse range of physical and social-cognitive abilities (measured using the Paediatric Evaluation of Disability Inventory). We benchmarked this analysis with the cut points approach calculated using the best available thresholds for the population. Time spent active as measured by this unsupervised approach correlated more strongly with measures of the childs mobility, social-cognitive capacity, independence, daily activity, and age than that measured using the cut points approach. Unsupervised machine learning offers the potential to provide a more sensitive, appropriate, and cost-effective approach to quantifying physical activity behaviour in diverse populations, compared to the current cut points approach. This, in turn, supports research that is more inclusive of diverse or rapidly changing populations.
Subjects: Quantitative Methods (q-bio.QM)
Cite as: arXiv:2201.10322 [q-bio.QM]
  (or arXiv:2201.10322v2 [q-bio.QM] for this version)

Submission history

From: Christopher Thornton PhD [view email]
[v1] Tue, 25 Jan 2022 13:50:30 GMT (1172kb)
[v2] Sat, 19 Feb 2022 10:07:23 GMT (1173kb)

Link back to: arXiv, form interface, contact.