By developing complex data models for yield prediction and employing portfolio optimisation theory for seed variety selection, we can increase the yield for up to 25%.
The underlining principle of precision agriculture is “producing more with less”. Research at BioSense Institute aspires to maximise crop yield using only the resources already available to farmers. By using big data analytics, namely the classifier based on the weighted histograms of yield, we have developed a statistical model of seed varieties. This model takes into account soil type, pH level and GPS coordinates and creates recommendations to a farmer which seed varieties to plant in order to maximise the yield. And the BioSense’s model is pretty good at it. It can predict the yield with an accuracy of ±15%.
After making predictions, it is time to pick the best varieties for a particular farm. Surely, we could just pick the most promising variety and plant it on the whole field, but there are smarter solutions. Some seeds are proper record breakers, but only in perfect weather conditions. In the case of drought or a disease outbreak, the harvest is going to be rather poor. Others are, however, more resilient to drought or pests, but have a smaller average yield. The catch is to find their winning combination. In order to find the right answer, BioSense’s data scientists had to dig deeper in yet another area – economics. We discovered that the portfolio optimisation theory, originally used for choosing the right investment at a stock exchange, can solve the problem. The main difference is that instead of monetary assets we are aiming to maximise crop yield and that the risk is not associated with stock-market fluctuations, but rather with the unpredictability of the weather.
All in all, using prediction and selection methods developed at BioSense, we managed to improve the yield in as much as 75% of cases! We are proud to say that using our technology, farmers could expect the yield increase of up to 25%, or additional earnings of up to 400€ per hectare of soybeans.