Dimensional Reduction Techniques for the Characterization of Behavioral Patterns in Dairy Cows

The current global population growth has intensified the demand for feed, especially dairy and meat products. In response to this growing need, precision livestock farming has become essential to manage animals efficiently and sustainably. Within this new approach, individual monitoring of each animal is essential. To this end, activity monitoring collars have been postulated as one of the most valuable tools for this purpose. With this in mind, three dimensional reduction methods (ICA, Isomap and t-SNE) have been analyzed in this study to identify patterns in the daily behavior of dairy cows through data collected by these collars. T-SNE was found to be especially effective in distinguishing individual or small group behaviors of animals, which is crucial for improving herd management and early detection of health or behavioral problems.