Monitoring behavior and its related energy expenditure of a desert goat through inertial and heart rate sensors

Monitoring behavior of grazing animals offers great potential to improve livestock management. However, few technologies are capable of collecting data and a limited number of methods may be used to classify these data and determine behavior. We propose an inertial measurement unit (INR) for classifying the behavior and heart rate (HR) monitoring to estimate energy expenditure (EE) of one (a=1) goat grazing in the Argentinean Monte Desert. Behavior was classified as resting in the pen (RP), resting in the field (RI), walking (W) and grazing (G). The behavior classification was performed using the Random Forests algorithm. Walking was detected with the highest precision (96%), while the precision was lowest for Grazing (80%). Goat activities could be predicted with an average precision above 86% and a recall of 85%, which suggests viability for real-life applications. Total EE of grazing was 535.1 KJ ME kg BW(0.75)d(-1), while EE of W, G, RF and RP were 53.8,108.5, 99.0 and 273.9 K3 kg BW0.75 d(-1), respectively. The goat walked 7,187 km, at an average speed of 4.22 km h(-1), with an energy cost of displacement at 7.48 J kg BW0.75 m(-1). These results expose the viability, relevance and usefulness of applying this methodology to study different factors affecting the behavior and energy expenditure of goats grazing in arid areas.