Applicability of Machine Learning to Improve Mastitis Prediction in Livestock

In recent years, the automation of dairy management using artificial intelligence has been sought after, with mastitis detection being one such application. Mastitis is an inflammatory response that occurs in dairy cows, leading to economic losses such as decreased milk production. Therefore, early detection is desirable. Currently, raw milk analysis devices using lactate dehydrogenase (LDH), a biomarker, are widely used for early mastitis detection. However, the use of sensor systems often results in false positives. It is common practice to refer to detection results from the past few days for the final infection judgment, which relies on the farmer’s experience, leaving room for improvement. This study aims to combine raw milk analysis devices with machine learning techniques to detect mastitis more accurately without relying on the farmer’s experience. We constructed three machine learning detection models, achieving a maximum recall of 0.89, precision of 0.81. Furthermore, the infection prediction approach proposed in this study is widely applicable and can achieve more advanced predictions when combined with related research.