Multivariate time series analysis with recurrent neural networks for dairy cattle breeding management

Negative scenarios can occur during calving of dairy cattle with high economic value. Many negative scenarios such as calf entrapment, injury, losses due to malnutrition after calving and unattended calving can occur. In order to reduce these adverse scenarios and improve animal welfare, accurate prediction of calving time is of great importance. In this study, we propose a new method for calving time estimation using a Bi-LSTM classifier with Kalman filter. The Bi-LSTM classifier is a deep learning model that can learn long-term dependencies in time series data. The Kalman filter is a statistical filter that can be used to smooth noisy data and estimate the state of a system over time. As a result of the experiments, it is found that the proposed model outperforms other alternative models and other forecasting methods in the literature. In particular, the Kalman filter provides an average of 20% better performance in the models, while the model achieves a high success with 93% precision and 95% accuracy. In addition, the cross-validation method reduced the effect of data set size and prevented overfitting, resulting in more accurate model results. Our method can be a valuable tool for dairy farmers. It can help farmers make more informed decisions about herd management, which can lead to increased animal welfare and milk production. This thesis demonstrates that artificial intelligence methods have a significant potential for application in the field of animal farming.