Exploring Milk Productivity and Energy Output Dynamics across Lactation Periods in Automatic Milking Systems through Multi-Algorithm Clustering

Automatic Milking Systems (AMSs) are pivotal in Precision Dairy Farming. These systems collect extensive data on cow level, milk production, and herd management at each milking event. Machine Learning (ML) methods are well-suited for analyzing the large amount of data generated by AMSs, providing tools to support herd management data-driven decisions. One critical decision in dairy farming is whether to retain a cow for future lactation periods, a choice made at insemination time but whose outcome is observed only in subsequent lactation cycles. ML can assist decision-makers by identifying cows with potential for maintaining or increasing milk production. This study applies Multi-Algorithm Clustering Analysis to assess the continuity of Productivity Groups (PGs), defined by Low and High milk yield and quality, over seven lactation periods across 16 AMS farms. Farms with more stable PGs were able to retain more cows for future lactation periods, although higher daily milk production was observed in farms in which Low PGs increased over time. In contrast, stable PG farms exhibited higher protein content across all lactation periods, with higher Fat-Protein Corrected Milk in the third lactation. Significant differences between stable and non-stable PG farms were primarily observed in the first lactation, when farms with non-stable PGs had smaller delivery ages and refusals by day, but higher milkings by day and days in milk. These findings suggest that cows on these farms may experience increased pressure during the first lactation, potentially leading to a greater decline in productivity and herd size over subsequent lactation periods.