RESUMO
Mastitis is one of the major health problems in dairy herds leading to a reduction in the leading to a reduction in the quality of milk and economic losses. The research aimed to present the system, which uses electronic 3D motion detectors to detect the early symptoms of mastitis. The system would allow more effective prevention of this illness. The experiment was carried out on 118 cows (64 Holstein Friesian and 54 Brown Swiss). The animals were kept in free-stall barn with access to pasture. The occurrence of mastitis cases was noticed in veterinary register. Microbiological culture was taken from milk in order to confirm the development of infection. Data from motion detectors were defined as time spent by animals on feed intake, ruminating, physical activity and rest, and were expanded by adding information about feeding group, breed type and lactation number. During analyses, two approaches were used to process the same dataset: artificial neural networks (ANN) and logistic regression. The obtained ANN and the logistic regression models proved to be satisfactory from the perspective of applied criteria of goodness of fit (area under curve-exceed 0.8). Quality parameters (accuracy, sensitivity and specifity) of logistic regression are relatively high (larger than 0.73), whereas the ranks of significance of the studied variables varied across datasets. These proposed models can be useful for automating the detection of mastitis once integrated into the farm's IT system.