Technical note: Mining data from on-farm electronic equipment to identify the time dairy cows spend away from the pen

Publication date: Available online 17 March 2017
Source:Journal of Dairy Science
Author(s): A.J. Thompson, D.M. Weary, M.A.G. von Keyserlingk
The electronic equipment used on farms can be creatively co-opted to collect data for which it was not originally designed. In the current study, we describe 2 novel algorithms that harvest data from electronic feeding equipment and data loggers used to record standing and lying behavior, to estimate the time that dairy cows spend away from their pen to be milked. Our 2 objectives were to (1) measure the ability of the first algorithm to estimate the time cows spend away from the pen as a group and (2) determine the capability of a second algorithm to estimate the time it takes for individual cows to return to their pen after being milked. To achieve these objectives, we conducted 2 separate experiments: first, to estimate group time away, the feeding behavior of 1 pen of 20 Holstein cows was monitored electronically for 1 mo; second, to measure individual latency to return to the pen, feeding and lying behavior of 12 healthy Holstein cows was monitored electronically from parturition to 21 d in milk. For both experiments, we monitored the time each individual cow exited the pen before each milking and when she returned to the pen after milking using video recordings. Estimates generated by our algorithms were then compared with the times captured from the video recordings. Our first algorithm provided reliable pen-based estimates for the minimum time cows spent away from the pen to be milked in the morning [coefficient of determination (R²) = 0.92] and afternoon (R² = 0.96). The second algorithm was able to estimate of the time it took for individual cows to return to the pen after being milked in the morning (R² = 0.98), but less so in the afternoon (R² = 0.67). This study illustrates how data from electronic systems used to assess feeding and lying behavior can be mined to estimate novel measures. New work is now required to improve the estimates of our algorithm for individuals, for example by adding data from other electronic monitoring systems on the farm.



http://ift.tt/2m9ck4Q