Good harvest forecasts are key for an optimal pricing strategy and labor planning. Under and over production, for example, can have major consequences for pricing. In addition, you can only deploy your manpower properly if you know how much and when they should harvest. Making accurate predictions about the amount of harvestable product for the coming weeks is a complex process and remains a big challenge. To improve these predictions, major steps are currently being taken with Artificial Intelligence (AI).
With Machine Learning, Hoogendoorn developed an AI-model that is capable to provide predictions for one to four weeks in advance with an accuracy up to 93% on average. For this the model uses both historical and expected data, for example about the greenhouse temperature. The first models were developed in 2017 and customers are now using the harvest forecast module at dozens of locations.
Despite the good results, continuous improvements are being made to achieve even better predictions. A team of data specialists is constantly working to optimize the underlying algorithms of the harvest forecast models, both in collaboration with customers and independently. The major advantage of using Machine Learning is that larger amounts of data almost automatically lead to more accurate forecasts. This leads to more accurate results than can be obtained by manually prepared predictions.
An AI model always keeps learning. When you feed the model more information, its understanding of the information will get better and better. This has a positive effect on the correlations the model makes. However, the model can only generate correct predictions when it receives correct information. The quality of data constantly improves due to technological developments, such as automated counting and harvesting. In addition, the harvest forecast model flawlessly works together with the iSii process computer. The iSii delivers more accurate data, because it makes use of sensors. The required data is entered into the model directly from the iSii. This saves time in two areas: when data is delivered and when calculations are made.
For more information:
Hoogendoorn Growth Management
info@hoogendoorn.nl
www.hoogendoorn.nl