Unsupervised machine learning for developing personalised behaviour models using activity data
- Submitting institution
-
University of the West of England, Bristol
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 887981
- Type
- D - Journal article
- DOI
-
10.3390/s17051034
- Title of journal
- Sensors
- Article number
- 1034
- First page
- -
- Volume
- 17
- Issue
- 5
- ISSN
- 1424-8220
- Open access status
- Compliant
- Month of publication
- May
- Year of publication
- 2017
- URL
-
http://www.mdpi.com/1424-8220/17/5/1034
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
4
- Research group(s)
-
-
- Citation count
- 10
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This paper utilises real-world data to illustrate the efficacy of using unsupervised machine learning (namely K-Means clustering and Self-Organising Maps) to model people’s behaviour patterns using low-cost commercial sensors. The models are validated using a ‘blind’ approach with clustering of clinically similar participant exceeding 85% when compared to carers’ clinical classifications. The significance of this research study is that it provides evidence for deploying this approach at scale, addressing the constraints of annotating numerous sensor data streams for activity classification, therefore making this a pragmatic solution for care providers to monitor and track peoples’ care needs over time.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
