Validation of automatic systems for monitoring the licking behaviour in Angus and Brahman cattle

Abstract

<p>Monitoring feeding time in ruminants is one means to quantify feed intake. In grazing cattle offered feed supplement blocks, time spent licking can provide valuable information in estimating the level of blocks being ingested. This current study aimed to 1) assess an ear-tag accelerometer's capability to identify the licking behaviour at supplement blocks in grazing cattle and 2) evaluate the performance of the ear-tag accelerometer and radio-frequency identification (RFID) system to predict the individual time spent licking. Two breed groups of Angus (<i>n</i> = 7) and Brahman (<i>n</i> = 7) beef heifers were kept in two separate yards over 28 days. Each heifer was fitted with an ear-tag containing a tri-axial accelerometer set at 12.5 Hz frequency. Feed supplement blocks were provided through an RFID-equipped automatic supplement weighing unit within each yard, with access to the unit being given daily only from 16:00 h - 20:00 h. The accelerometer classification model developed using support vector machine (SVM) algorithm could distinguish between licking and non-licking behaviours, with an accuracy, sensitivity, F1 score, Cohen's kappa coefficient, and Matthew's correlation coefficient (MCC) of 86%, 93%, 0.88, 0.70, and 0.77 for Angus and 87%, 93%, 0.89, 0.73, and 0.79 for Brahman heifers. Time spent licking predicted by accelerometers were acceptable with a mean absolute error (MAE) of 22% and 11%, modelling efficiency (MEF) of 0.81 and 0.94, concordance correlation coefficient (CCC) of 0.88 and 0.96, and a ratio of root mean square prediction error (RSR) of 0.44 and 0.25, for Angus and Brahman heifers, respectively. However, the RFID system derived predictions of time spent licking in grazing heifers were unacceptable for both breeds. Overall, the ear-tag accelerometer offers the potential to predict individual time spent licking in grazing cattle to estimate block supplement intake.</p>