Description

Four models were developed for predicting energy expenditure in preschoolers using raw acceleration features from either a hip or non-dominant wrist-worn ActiGraph accelerometer. Models were trained using 20-min of free-play from 15 participants and cross-validated in a holdout sample of 10 participants.

The following models are available. Root Mean Square Error (RMSE) is also provided.

• Random forest for the non-dominant wrist (0.96 METs)

• Support vector machine for the non-dominant wrist (0.99 METs)

• Artificial neural network for the right hip (0.96 METs)

• Random forest for the right hip (0.96 METs)

Development/Validation

Sample: 25 (20 males) healthy children, 3-5 years of age

Setting: Home, park, other community setting

Activities: 20-min of free-play

Criterion: MetaMax 3b (VO₂)

Accelerometer(s): ActiGraph GT3X+ on hip or non-dominant wrist

Validation approach: Holdout of 10 participants

Phase Designation

(What's this?)

This model is in Phase 2 and 3.

Instructions

The R code can be used to read in a 'csv' file with the necessary features and then use the chosen model to predict energy expenditure in kcals/min for each 10-sec window. There is a separate file for each of the four models- the wrist random forest, wrist support vector machine, hip artificial neural network, and hip random forest. For both the hip and wrist there is an example data sheet with the variables described in the data dictionary that need to be present in the csv file. Further instructions are in the R code. More information about R is found here.

This code relies on R software which can be downloaded for free at https://www.r-project.org/

Attached Files

Download all files as a .zip or download individual files below.

Code

Example hip data

Example wrist data

Data dictionary

Wrist random forest

Wrist support vector machine

Hip artificial neural network

Hip random forest

Note: Compared to the original source, code and data dictionary were edited for consistency, example data were added, models were unchanged

Source Information

Reference:

Ahmadi MN, Chowdhury A, Pavey T,Trost SG (2020) Laboratory-based and free-living algorithms for energy expenditure estimation in preschool children: A free-living evaluation. PLoS One 15(5): e0233229. https://doi.org/10.1371/journal.pone.0233229 Link to paper

Original source for code:

https://github.com/MA-QUT/Preschool_EE_Models_PLOS_One

Corresponding author:

Stewart G. Trost, s.trost@qut.edu.au

Contact

Kimberly Clevenger at accelerometerrepository@gmail.com