Automated detection of behaviour in Caucasian Tur

Description

One female Caucasian Tur (Capra caucasica) was equipped with a collar that detected head movement using acceleration sensors in 3D. The animal was observed for up to two hours per day for several months and the noted behaviour was correlated with the sensor data in order to train an algorithm for automated behaviour detection. The analyses were done in R with the Random Forrest algorithm. The results show that over 20 different classes of behaviour (including head postures, rumination, different locomotion types) can be identified with very high accuracy (>95%).

Location