Model of neural networks for fertilizer recommendation and amendments in pasture crops

In the production of milk, pastures are the basis of bovine feeding, it is considered that around 50% of the costs to produce a liter of milk correspond to the feeding of the cows. Therefore the cultivation of pastures is an essential factor in the dairy industry. The cultivation of pastures requires the preparation of the soil adding fertilizers and amendments for the increase of biomass. The determination of fertilizers and amendments is an activity that adds costs to the cultivation process. The purpose of this work is to identify that from the basic nutrients in the soil such as Nitrogen(N), Phosphorus(P) and Potassium(K), with an MLP neural network of multiple input layers and multiple output layers trained with the back propagation algorithm, it can determine the fertilizers and amendments required by pasture cultivation. The results indicate that despite using a small sample of data (44), with a threshold of 0.75 and with k-fold =3 it was possible to identify the fertilizers Potassium chloride(Y2), Diammonium phosphate(Y6), Copper sulfate(W3) and Zinc sulfate(W4) for the cultivation of pastures. With these results, our contribution is to initiate the use of ML in pasture cultivation and later extend it to other crops. The importance of this work lies in the fact that the results can contribute to the reduction of milk production costs and that the results of this work will be the baseline for the use of other machine learning algorithms in agriculture.