Use of an ear-tag accelerometer and a radio-frequency identification (RFID) system for monitoring the licking behaviour in grazing cattle

Abstract

Monitoring the licking behaviour in grazing cattle is a potential means for quantifying block supplement intake. The current study aimed to 1) evaluate the capability of an ear-tag accelerometer to identify the licking behaviour at a block supplement in grazing cattle and 2) validate individual licking state (LS) duration predicted by an accelerometer and radio-frequency identification (RFID) system. Four out of 12 Angus steers weighing 384 ± 9.7 kg (mean ± SD) were given free access to a 900 m2 supplement yard with access to two RFID-equipped automatic supplement feeders provided daily from 15:00 to 18:00 h for 10 days. Each steer was fitted with an ear-tag containing a 3-axis accelerometer set at a frequency of 25 Hz. Accelerometer data were segmented into three window sizes (3, 5, and 10 s) and further processed using four machine learning (ML) algorithms: Random Forest (RF), Extreme Gradient Boosting (XGB), Logistic Regression (LR) and Linear Discriminant Analysis (LDA). The best performance in classifying licking behaviour was obtained from the combination of XGB and 10 s window size with an accuracy, Kappa coefficient, and F1 score of 93%, 0.88, and 0.88, respectively. Accelerometers and RFID systems consecutively under-predicted and over-predicted LS duration by 20% and 6%, with a mean absolute error (MAE) proportion of 22% and 10%, a ratio of root mean square prediction error (RSR) of 0.33 and 0.14 and a modelling efficiency (MEF) of 0.89 and 0.98. Overall, both the ear-tag accelerometer and RFID system was capable of monitoring the licking behaviour and LS duration of grazing cattle accessing block supplements.