Description

Thirteen models were developed for predicting energy expenditure in children using ActiGraph monitors worn at the chest, and each hip, ankle, wrist, and knees (and combinations of monitors). Models were developed for count data using children playing active video games and walking/running on a treadmill and cross-validated using leave-one-out cross-validation.

The following models are available. Root mean square error (RMSE) is also provided.

• Artificial neural network for all nine accelerometers (1.46 METs)

• Artificial neural network for chest (1.2-1.3 METs*)

• Artificial neural network for chest, right hip, right ankle (1.2-1.3 METs*)

• Artificial neural network for chest, right knee, right wrist, right hip (1.2 METs*)

• Artificial neural network for left ankle (1.2-1.3 METs*)

• Artificial neural network for left hip (1.2-1.3 METs*)

• Artificial neural network for left knee (1.3 METs*)

• Artificial neural network for left wrist (1.3 METs*)

• Artificial neural network for right ankle (1.2-1.3 METs*)

• Artificial neural network for right hip (1.2-1.3 METs*)

• Artificial neural network for right knee (1.3 METs*)

• Artificial neural network for right wrist (1.3 METs*)

• Artificial neural network for right wrist, right knee (1.2-1.3 METs*)

*estimated from Figure

Development/Validation

Sample: 27 (15 males) healthy children with a mean age 10.8±1.1 y

Setting: Laboratory

Activities: Active video games, treadmill walking and running

Criterion: MetaMax 3b (VO₂)

Accelerometer(s): ActiGraph GT3X+ at the chest, both hips, wrists, ankles, knees (not all needed for models)

Validation approach: Leave-one-out cross-validation

Phase Designation

(What's this?)

This model is in Phase 1.

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 METs for 15-sec windows. There is a separate file for each of the thirteen models. There is an example data sheet with the variables described in the data dictionary. Since not all monitors are used in each model, you will only need the variables for the appropriate monitor, not every variable present in the example data sheet. 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 data

Data dictionary

Artificial neural network for all nine monitors

Artificial neural network for chest

Artificial neural network for chest, right hip, right ankle

Artificial neural network for chest, right knee, right wrist, right hip

Artificial neural network for left ankle

Artificial neural network for left hip

Artificial neural network for left knee

Artificial neural network for left wrist

Artificial neural network for right ankle

Artificial neural network for right hip

Artificial neural network for right knee

Artificial neural network for right wrist

Artificial neural network for right wrist, right knee

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

Source Information

Reference:

Mackintosh, K. A., Montoye, A. H. K., Pfeiffer, K. A., & McNarry, M. A. (2016). Investigating optimal accelerometer placement for energy expenditure prediction in children using a machine learning approach. Physiological Measurement, 37(10), 1728. https://doi.org/10.1088/0967-3334/37/10/1728 Link to paper

Original source for code:

https://sites.google.com/site/alexmontoye/machine-learning-model-code

Corresponding author: Kelly Mackintosh, k.mackintosh@swansea.ac.uk

Contact

Kimberly Clevenger at accelerometerrepository@gmail.com