Monitoring milking parameters to improve milking operations through machine learning algorithms

The operation of milking is one of the most time-consuming in a dairy cattle farm. Because the management and duration of the whole milking session can be affected by some cows that need a longer milking time than others, it can be useful to shorten the milking time of these cows. In this study, a full dataset of milking data was collected and processed for three months from a dairy cattle farm located in Northern Italy. The aim was to understand how to reduce the daily milking time by evaluating the effect of a different pulsation ratio and detachment flow rate on the duration of milking and udder health. A prediction model for the duration of milking was developed, which was able to identify the proper pulsation ratio and detachment flow rate based on the first 2 minutes of data on milking. If implemented on machines, it can lead to an automatization in the change the pulsation ratio and detachment flow of every cow.