Combined use of two supervised learning algorithms to model sea turtle behaviours from tri-axial acceleration data

Abstract

Accelerometers are becoming ever more important sensors in animal-attached technology, providing data that allow determination of body posture and movement and thereby helping to elucidate behaviour in animals that are difficult to observe. We sought to validate the identification of sea turtle behaviours from accelerometer signals by deploying tags on the carapace of a juvenile loggerhead (), an adult hawksbill () and an adult green turtle () at Aquarium La Rochelle, France. We recorded tri-axial acceleration at 50 Hz for each species for a full day while two fixed cameras recorded their behaviours. We identified behaviours from the acceleration data using two different supervised learning algorithms, Random Forest and Classification And Regression Tree (CART), treating the data from the adult animals as separate from the juvenile data. We achieved a global accuracy of 81.30% for the adult hawksbill and green turtle CART model and 71.63% for the juvenile loggerhead, identifying 10 and 12 different behaviours, respectively. Equivalent figures were 86.96% for the adult hawksbill and green turtle Random Forest model and 79.49% for the juvenile loggerhead, for the same behaviours. The use of Random Forest combined with CART algorithms allowed us to understand the decision rules implicated in behaviour discrimination, and thus remove or group together some 'confused' or under--represented behaviours in order to get the most accurate models. This study is the first to validate accelerometer data to identify turtle behaviours and the approach can now be tested on other captive sea turtle species.

Behavioural inference from signal processing using animal-borne multi-sensor loggers: a novel solution to extend the knowledge of sea turtle ecology

“…Indeed, the development of fast personal computers and of free user-friendly computing libraries made it possible to easily apply these 'black box' algorithms to huge amounts of data. The machine-learning approach has thus turned out to be a very powerful tool for identifying well-characterized behaviours (in terms of signal) such as locomotion [56][57][58] and resting [59][60][61]. However, it appears to be rather inefficient when seeking to identify behaviours with confusing signal characteristics.…”

Monitoring canid scent marking in space and time using a biologging and machine learning approach

“…3-dimensional) inertial sensors (see 41 and papers therein). The large datasets these high resolution sensors produce can be analysed using machine learning techniques to quickly and automatically find behaviours of interest [41][42][43][44] . One such machine learning technique, the k-Nearest Neighbour (KNN) algorithm, is a conceptually simple and effective method of classifying behaviours in accelerometer data 45 .…”

Augmenting biologging with supervised machine learning to study in situ behavior of the medusa Chrysaora fuscescens

“…As tethering can restrict movement, it is unknown whether data collected in this manner are representative of natural behavior. Furthermore, without simultaneous observation of the tagged animal, interpretation of biologging data can be easily biased (Brown et al, 2013; Jeantet et al, 2018).…”

“…Recently, techniques from supervised machine learning (ML), which automatically fit or ‘learn’ patterns that optimally distinguish categories, have been successfully used to classify behaviors in various marine vertebrates (Brewster et al, 2018; Jeantet et al, 2018; Ladds et al, 2016). However, few studies develop their methods on ground-truthed in situ data owing to the difficulty of recording sustained observations of wild marine animals (Carroll et al, 2014).…”