- An international animal health company wanted to build an improved decision tool for beef producers looking for the best time to market their cattle. This relied on, among other things, improving the accuracy of their current models to predict the endpoint performance of cattle by days on feed, using more advanced modeling methods and more granular data.
We applied the latest machine learning methods to:
- Predict five different performance metrics for each cattle lot, such as carcass weight and quality grade.
- Incorporate daily data on feed intake, treatments, mortality, weather, etc.
- Forecast key daily metrics, such as feed intake, out to the lot endpoint.
- The use of machine learning methods, in comparison to the traditional approach, improved upon the predictive accuracy by 10 to 20 percent. We also proposed a broader solution to be implemented, incorporating models for cattle price (revenue) and feed prices (cost) to build a recommendation / optimization engine.