Determination of Cattle Standing-Time with Decscion Trees and Neural Nets by using Only Acceleration Data from Collar

The method, data analysis, selection and use of different classifiers with the corresponding detection rates for the evaluation of data of an acceleration sensor on the collar on dairy cows for detecting standing-up and laying-down are presented in this paper. Data was collected in video surveilled measurement series. The studies are targeting the development of an activity monitoring system integrated in a collar without an additional ankle monitor. This paper will investigate the development of classifier by using different machine-learning techniques, the "coarse tree" combined with saviztzky-golay differentiation and a neural net with backpropagation algorithm. At the end the results of the methods used are presented and compared.