I’m not so much into hackathons anymore, but I’m always willing to make an exception for hacking farming data.
Lely and Rovecom joined forces to provide a unique dataset containing both cow diets and milk robot data. The diet set contained a lot of parameters on the cow’s input while the robot dataset a lot of parameters on the cow’s output. In total there were more than 2 500 cows in the combined dataset.
The challenge: provide insights into the relation between the cow’s input and output.
The challenge was very popular among the participants, probably thanks to the excellent dataset provided. We managed to form three teams of which two ended up among the (three) hackathon winners.
Team green: the challengers from Lely & Rovecom, the three participating teams and the mentors
The winner – team Betsie – was manned mostly by the founders of the Track32 company and one WUR student they adopted. They used linear regression to show clear relationships between certain diet ingredients and cow’s heath and milk quantity and quality. They also identified a number of missing variables, such as amount of grass/hay consumed. They proposed a creative way of collecting this data using cameras and image processing software. Not surprisingly, as image processing seems to be their core business :-). Cool stuff.
Team Betsie kicking off their presentation
A linear regression model showing the influence of the diet parameters (in blue) on milk production
The second runner up – team DairyBite – organized a mini-Kaggle competition to come up with the best predictive model. They also hacked together a cute Vue.js fronted app that really impressed the jury. And they designed a cool logo – definitely the best logo of this hackathon.
DairyBite – the best logo of this hackathon
The DairyBite frontend app
Hopefully these excellent results will help Rovecom and Lely to figure out even better ways of feeding cows and making them happier. And who knows, maybe that will make our farmers also happier.