In the animal feed industry, a lot of research is done to optimize the performance of dairy cattle. Through in vivo and in vitro tests performed by research facilities of feed companies, we start to get more insights in predicting the animal performance. This knowledge has the potential to enhance the productivity of the dairy cattle and increase revenue for the farmer.
With this research assignment, the student will explore the possibilities to use AI and machine learning as techniques to predict animal performance. We want to investigate whether there is a correlation between different input parameters, such as the composition of the ration and the nutritional quality of the ration, with the actual measured output on the farm. The prediction calculated via software might be very different from the actual situation. Today, figures on this topic are not known.
The student will formulate an answer to these questions:
- How can machine learning and AI help the industry/farmers to get a better prediction on these variables?
- Are these predictions better than via complex mathematical equations done today?
The starting point of this research is the current Dutch cow model. In short: when formulating a ration, several input parameters are used to calculate the animal requirements (such as: days into lactation, DMI, milk protein and fat, age of the animal, body weight, calving interval and location). With these input parameters, a ration is calculated using certain raw materials. The ration will have a certain quality in terms of level of protein, level of energy, amino-acids, etc. This will return a set of parameters as output of the ration: allowable milk from energy and protein, predicted DMI and roughage intake.