Estimativa da produtividade da soja com redes neurais artificiais

• Ano de defesa: 2016
• Autor(a) principal: Alves, Guiliano Rangel
• Orientador(a): Teixeira, Itamar Rosa
• Banca de defesa: Melo, Francisco Ramos de, Leandro, Wilson Mozena
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual de Goiás
• Programa de Pós-Graduação: Programa de Pós-Graduação Stricto sensu em Engenharia Agrícola
• Departamento: UEG ::Coordenação de Mestrado em Engenharia Agrícola
• País: Brasil
• Palavras-chave em Português: Glycine max; Características agronômicas; MLP; perceptron; Agronomic characteristics; Perceptron
• Palavras-chave em Inglês: Glycine max
• Área do conhecimento CNPq: CIENCIAS AGRARIAS::ENGENHARIA AGRICOLA
• Link de acesso: http://www.bdtd.ueg.br/tede/handle/tede/209

Nowadays, to estimate soybeans productivity are used complex statistical models, which turns limited the access to this practice. An alternative to these models is use of computer systems applying Artificial Intelligence (AI). On this line of system, an option is apply Artificial Neural Networks (ANN), which has the capacity of learning through examples of problems presented. This work aimed then: evaluate the possibility of using the Multilayer Perception (MLP) ANN to estimate the productivity of soybean based on the growing habits, seeding density and agronomical characteristics; define the relevant parameters during the ANN developing to evaluate the agronomical characteristics and its relation with soybean productivity; develop and choose an ANN architecture to solve the proposed problem. To realize the work were used agronomical data of the soybean culture obtained on experiments leaded on 2013/2014 harvest at Anapolis-GO, which data were normalized on compatible range to work with the ANN and then done the training of the several ANNs, to choose the ANN with best performance. After the network training were realized a performance analysis of each one to select the ANN with the most appropriate answer to the problem. The chosen ANN indicated a success range of 98% with the training data and 72% with the validation data. This can be considered a high level of achievement, mostly if considered the complexity of factors involved on the estimative of the soy productivity. The ANN application of the kind MLP on the conducted experiment data shows that is possible to estimate the soy productivity based on agronomical characteristics, growing habits and seeding density through AI.