Research and Classification of Eating Behavior Patterns of Cattle Using a Convolutional Neural Network

The study provides an overview of evidence on the use of feeding characteristics in cattle to identify health and welfare problems. A review of machine learning algorithms suitable for classifying the feeding behavior of cattle is presented and the use of a convolutional neural network (CNN) is justified. The use of intragastric bolus and undertightening as sources of input data for CNN is shown. To solve the problems of practical implementation of the developed CNN, a compromise between CNN performance, computational complexity and required hardware resources was identified, a search for hyperparameters was performed, and the CNN was optimized.