Krieter, JoachimStamer, EckhardJunge, WolfgangWenkel, K.-O.Wagner, P.Morgenstern, M.Luzi, K.Eisermann, P.2019-08-262019-08-2620063-88579-172-2https://dl.gi.de/handle/20.500.12116/24700Exponentially weighted moving average control charts and neural networks were used for oestrus detection in dairy cows. The analysis involved 373 cows, each with one verified oestrus event. Model inputs were the traits activity, measured by pedometer, and the period (days) since last oestrus. In total 10,386 records were available, which were partitioned into training and validation subsets to train and test the neural network (multifold cross-validation). When the trained neural network was applied to the validation sets, the averaged sensitivity, specificity and error rate were 77.5, 99.6 and 9.1%, respectively. Performance for the same data with the univariate control chart was less successful. Neural networks are useful tools to improve computerised oestrus detection in dairy cows.enText/Conference Paper1617-5468