Análise e modelagem estatística da produtividade da cana de açúcar usando machine learning

Abstract

Sugarcane is mainly used for the production of sugar and renewable bioenergy. Estimating sugarcane productivity in a precise and timely manner plays a particularly important role in agribusiness management, requiring quick and accurate responses, since there are numerous factors that determine productivity and they are constantly changing. The objective of this dissertation is to generate predictive models for sugarcane productivity that present precision above the resources currently available in the market, through the association of statistical methods and artificial intelligence in the strategic management and optimization of the productive process. The research evaluated data on agricultural practices related to management and edaphoclimatic factors of crops with sugar cane. The regression models performed predictions for new data with 96.44% accuracy for the PLS regression model and 99.97% accuracy using machine learning techniques