Feed intake is very important for dairy cows and deviation from normal eating behaviour may predict a cow that needs treatment. We used video recordings of dairy cows at the Danish Cattle Research Centre (DKC) combined with data from a neck-collar mounted 3D accelerometer and RFID device from Lyngsoe Systems (Aars, Denmark) to develop a random forests model for predicting daily eating activity. We investigated performance by internal cross-validation and the results indicate that we obtain accurate predictions of daily eating time by the algorithm. Technical challenges are delaying the planned tests on commercial farms. We are therefore currently utilising historical data from DKC to examine the potential of using changes in daily eating time for detection of sick cows.
Original language
English
Title of host publication
Symposium i Anvendt Statistik
Editors
Peter Linde
Publication year
Jan 2017
Pages
109-122
ISBN (print)
978-87-501-2267-8
Publication status
Published - Jan 2017
Event
Symposium i Anvendt Statistik - Syddansk Universitet, Odense, Denmark Duration: 23 Jan 2017 → 24 Jan 2017 Conference number: 39
