Development of diagnostic, predictive and prescriptive tools based on milk components

A multitude of studies and databases that assemble a body of knowledge, expertise and practices in the dairy field have been accumulated over more than a hundred years. This expertise is used by researchers, agronomists and veterinarians to provide recommendations to dairy producers. However, with all the parameters associated with a farm and all the indicators to be taken into account, it is difficult, if not impossible, for these experts to continue to provide reliable recommendations without having access to tools that can amalgamate this information and quickly detect problems on the farm, such as the appearance of mastitis or subclinical acidosis. In this project, we are paving the way to build prescription tools in the dairy sector by proposing the algorithms and artificial intelligence methods needed to answer the four questions leading to prescription in the sector: a) What is happening on the farm from tank milk and individual cow data; b) Why is it happening on the farm; c) What is likely to happen in the future; d) How to prevent/avoid a risk and correct a situation.

The tools implemented will allow the exploitation of a knowledge graph capturing the knowledge space of practices, procedures and properties of the field from published articles, meta-analysis results, documented protocols and databases. They will also allow the construction of stimulators based on fuzzy approaches to establish metrics and probabilities, and to evaluate future properties. All these tools will lead to a prototype prescription system that will take the dairy domain to the cutting edge of symbolic approaches in artificial intelligence, capturing and making available the most beneficial practices and knowledge for the sector. These tools could be generalized to other properties or diseases requiring a prescription.