Análise comparativa de métodos de boosting aplicados à recomendação de cultivos agrícolas

Abstract

ABSTRACT

The use of predictive technologies, such as machine learning algorithms, emerges as a viable and efficient alternative to guide crop choices and recommendations. These algorithms improve the accuracy of agronomic decisions and mitigate environmental and economic risks. In this context, this work aimed to compare the performance of the XGBoost (eXtreme Gradient Boosting) and CATBoost (Categorical Boosting) and ADABoost (Adaptive Boosting) models in crop recommendation. A database with 6 different crops, 2 physical parameters, and 9 nutrients was used for this comparison. The XGBoost, CATBoost, and ADABoost models had their hyperparameters optimized using the Optuna library and were evaluated using the Accuracy, Kappa, and f1-score metrics. The results indicate that all three models performed very well in the task of crop recommendation, with high accuracy, Kappa, and f1-score values.