Journal: bioRxiv
Article Title: Leveraging machine learning and accelerometry to classify animal behaviours with uncertainty
doi: 10.1101/2024.12.28.630628
Figure Lengend Snippet: Smoothed classifications for a sample 12-hour African wild dog accelerometer segment. The top plot displays the original accelerometer signal, with shaded panels representing known labelled behaviours. The subsequent three plots illustrate the predicted probabilities of the different behavioural classes with no ( s = 1), moderate ( s =25), and high ( s =100) average window lengths.
Article Snippet: From October 2021 to September 2023, we deployed wildlife tracking collars with inbuilt GPS and accelerometer sensors (GPS PLUS 1C, Vectronic Aerospace, Germany) on five African wild dogs ( Lycaon pictus ) in the Okavango Delta, Botswana (centre 19 ◦ 31 S, 23 ◦ 37 E ).
Techniques: