Quantifying finer-scale behaviours using self-organising maps (SOMs) to link accelerometery signatures with behavioural patterns in free-roaming terrestrial animals.

These files are the supplementary data and code files for the following paper

Quantifying finer-scale behaviours using self-organising maps (SOMs) to link accelerometery signatures with behavioural patterns in free-roaming terrestrial animals.

Authors:

Nicole Galea, Fern Murphy, Joshua L. Gaschk, David S. Schoeman, Christofer J. Clemente

Corresponding Author: C. J. Clemente

Email: cclement@usc.edu.au

Description.

There are three data sets included in this set

Matlab files: Cats_align_video_6.m (+ Cats_align_video_6.fig)

These files are to be run in the matlab environment, and are primarily used to synchronise video and accelerometer data, and then tag individual behaviours. The input of this code is an accelerometer trace and a video file, which is synchronised using several taps. Instructions to use the code is provided in the supplementary material.

The output of this code is a data frame consisting of the raw X,Y,Z accelerometer data, trimmed to the length of an input video, and a column of tagged behaviour.

This output file is then input into the R environment using the following code.

R: Cat_code_maple_workshop.R

This code takes the output of the matlab script (can be several files which are read in a for loop) and applies the classifiers based on Tatler et al. (2018) applied in rolling epochs 50 samples long.

The output from this file is a combined file where each row represents a single epoch defined by 26 classifiers, and labelled with the most common behaviour during the epoch. This file is then used as the input for the SOM network.

Tatler, J., Cassey, P., & Prowse, T,A,A. (2018) High accuracy at low frequency: detailed behavioural classification from accelerometer data, Journal of Experimental Biology, 22(23): jeb.184085.

R: Cat_SOM_2020.R

This code reads in the output from the code above. It contains two functions:

trSamp – creates a training or testing data set with a variable sample size.

doSOM – calls up the trSamp function, runs the SOM, creates a confusion matrix and outputs the sensitivity, precision, specificity and overall accuracy.

The code also contains several lines for plotting and visualising the SOM output.