Skip to navigation Skip to main content Skip to footer

Approved research

Identifying patterns in accelerometer data, and investigating their association with other factors

Principal Investigator: Dr Louise Millard
Approved Research ID: 17810
Approval date: May 9th 2016

Lay summary

Typically, analyses of the association of physical activity with other traits have used a derived variable to represent the activity of individuals, such as moderate to vigorous physical activity (MVPA) and average counts per minute (CPM). There is likely to be much additional information in the accelerometer data, from which these measures are derived, which is not captured in these derived variables. We aim to use data mining methods to identify potentially interesting patterns in the accelerometer data, and then identify associations of these patterns with other traits. As use of the raw accelerometer data is not commonplace in epidemiological analysis, an important part of this project is the identification and development of appropriate methods for this purpose. Physical activity is associated with a wide range of traits and diseases. This project aims to improve understanding of the relationship between particular types of physical activity and other traits, by identifying and investigating patterns within the raw accelerometer data. It is important to identify patterns of activity associated with other traits in order to inform policy on the types of activity associated (perhaps causally) with other traits and diseases. This project will have two stages. The first stage will involve using machine learning methods to identify patterns in the accelerometer data. The second stage will test the association of the frequency of these patterns in participant?s accelerometer data with other traits. All participants with accelerometer data.