Methods for Classifying Physical Activities Using Accelerometer Data: A Scoping Review
Bibliographic record
Abstract
<title>Abstract</title> Accurate classification of physical activity from accelerometer data is critical for health research and large-scale population studies. However, the wide variety of computational methods used to derive activity types and intensity levels has led to inconsistencies in implementation, validation, and reproducibility. This scoping review aimed to identify and categorise methods used to classify physical activities using accelerometer data, with a particular focus on implementation, simplicity, validation strategies, and feasibility for application in large-scale datasets such as the All of Us Research Program. We searched PubMed, Web of Science, and SPORTDiscus for studies published between 2015 and 2025. Studies were included if they used accelerometer data to classify specific activities or general activity levels and reported a validation strategy. A total of 1,670 records were screened; 158 met the inclusion criteria. Data were extracted on study characteristics, classification methods, whether validation was performed, device use, specifications, and tool availability. Machine-learning techniques were the most frequently applied classification method (n = 81), followed by deep learning (n = 63), hybrid models (n = 23), rule-based or threshold approaches (n = 22), and unsupervised or other novel methods (n = 12). Walking (n = 97), sitting (n = 79), and standing (n = 68) were the most commonly studied activities. Most studies employed lab-based protocols and used k-fold or leave-one-subject-out validation. Only 16 studies provided public code or tools, and just a couple (n = 2) considered seasonality or population diversity. This review highlights substantial variation in activity classification methods and reporting practices. Open-source tool availability and validation in real-world conditions remain limited. There is a need for simpler, validated, and reproducible approaches, particularly for use in population-scale datasets like All of Us and the UK Biobank.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".