Modelagem de dados introdutória para predição do estado dos grãos de trigo através de análises multiespectrais

Abstract

This work aimed at an introductory study applied in agriculture, concerning data modeling to predict the condition of wheat grains through multispectral analysis of the AS7265x sensor. Ten commercial samples and 1 healthy sample of wheat grains were analyzed. Among commercial samples, 8 of them are contaminated with deoxynivalenol (DON). Fifty readings were collected from each typifying, totaling 550 data inputs. The paper does a performance analysis of supervised machine learning algorithms: k- nearest neighbors, support vector machine and random forest. As a result, the random forest algorithm had the highest performance among the 3. From the results, we conclude that this solution is a first step to help a wheat farmer in the decision-making process.