Predição da área foliar em acerola por redes neurais e regressão múltipla

• Autor(a) principal: Alcinei Místico de Azevedo
• Data de Publicação: 2019
• Outros Autores: Vitor Alves da Silveira, Celso Mattes Oliveira, Carlos Enrrik Pedrosa, Vinícius Teixeira Lemos, Nermy Ribeiro Valadares, Amanda Gonçalves Guimarães
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Repositório Institucional da UFMG
• Texto Completo: https://doi.org/10.32406/v2n32019/96-105/; http://hdl.handle.net/1843/40500; https://orcid.org/0000-0001-5196-0851
• topic: Acerola; Perceptrons; Redes neurais (Computação); Inteligência artificial

The objective of this work was to predict the leaf area in acerola by means of artificial neural networks (ANNs) and verify the efficiency of this methodology in comparison to multiple regression models. The length, width and area of 350 leaves of acerola were evaluated, 14 models of multiple regression and model of multilayer perceptron type RNA were used to predict the leaf area. The quality of fit between the multiple regression models and the ANNs was close, but the artificial neural networks were more efficient in the prediction of the leaf area in acerola, with determination coefficient superior to 0,98, being the network with two neurons in the intermediate layer the best prediction.