Random forest na previsão da produção de cana-de-açúcar
Carregando...
Data
2023-01-23
Autores
Ferreira, Hugo Guiné Pinto
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A cana-de-açúcar é a principal fonte de energia renovável no Brasil com um futuro promissor em todo o mundo, tanto na questão econômica quanto na ambiental. Assim, buscar métodos que melhorem a capacidade de previsão do rendimento da cultura é estratégico para a sustentabilidade da produção. O objetivo do trabalho foi testar o algoritmo de aprendizado de máquina Random Forest (RF) para previsão da produtividade da cana-de-açúcar a partir dos atributos do solo e práticas de manejos da cultura em áreas comerciais no noroeste paulista. Foram utilizados dados dos anos de 2016 a 2018, com um total de 70 mil hectares de área plantada. Os resultados indicam que a acurácia dos modelos encontrados estão entre 77 e 94%. As variáveis de manejo, tais como, número de cortes e época de colheita, foram as principais condicionantes da produtividade da cultura, em relação às variáveis químicas do solo. Os resultados indicam que o algoritmo RF apresentou precisão satisfatória, evidenciando sua aplicação na tomada de decisão dentro das unidades produtoras.
Sugarcane is the main source of renewable energy in Brazil with a promising future throughout the world, both economically and environmentally. Thus, the forecast of crop yield is strategic for the sustainability of production. The objective of this work was to test the Random Forest (RF) machine learning algorithm to predict sugarcane productivity based on soil attributes and crop management practices in commercial areas in northwest São Paulo, Brazil. Data from the years 2016 to 2018 were used, with a total of 70 thousand hectares of planted area. The results indicate that the accuracy of the models found were between 77 and 94%. Management variables such as number of harvests and harvest season were the main determinants of crop yield to soil chemical variables. Our results show that the RF algorithm was enough to be applied for decision making of sugar mill operation planning.
Sugarcane is the main source of renewable energy in Brazil with a promising future throughout the world, both economically and environmentally. Thus, the forecast of crop yield is strategic for the sustainability of production. The objective of this work was to test the Random Forest (RF) machine learning algorithm to predict sugarcane productivity based on soil attributes and crop management practices in commercial areas in northwest São Paulo, Brazil. Data from the years 2016 to 2018 were used, with a total of 70 thousand hectares of planted area. The results indicate that the accuracy of the models found were between 77 and 94%. Management variables such as number of harvests and harvest season were the main determinants of crop yield to soil chemical variables. Our results show that the RF algorithm was enough to be applied for decision making of sugar mill operation planning.
Descrição
Palavras-chave
Metodo da árvore de decisão, Agricultura de precisão, Big data, Precision agriculture DLC, Decision tree