Big data for cow disease prediction

The project’s aim was to create a reliable alarm system for detection of diseases in dairy cows, using large quantities of automatically collected milking data combined with manual registrations of multiple types of diseases. Earlier disease detection means better production, better animal welfare, more efficient treatment, and reduced antibiotics usage. Large quantities of data are collected during milking, however these are still not used optimally. In this project we had access to three large sets of high-quality cow-level data from the Netherlands, USA, and Denmark. The goal was to create dynamic models to describe data collected from healthy cows; use machine learning on the forecast errors of these models to predict specific problems and compare the performance of various methods. Our models were implemented on existing milking systems used in industry.

Related publications:

• M. van der Voort, D. Jensen, C. Kamphuis, I. N. Athanasiadis, A. De Vries, H.Hogeveen, Invited review: Toward a common language in data-driven mastitis detection research, Journal of Diary Science, 104(10):10449-10461, 2021, doi:10.3168/jds.2021-20311.