Aplicações de redes neurais e neuro fuzzy em engenharia biomédica e agronomia

• Ano de defesa: 2014
• Autor(a) principal: Silva, Inara Aparecida Ferrer [UNESP]
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Lógica difusa; Hemiplegia; Redes neurais (Computação); Fuzzy logic
• Link de acesso: http://hdl.handle.net/11449/110516

The fuzzy and neuro fuzzy systems have been successfully used to solve problems in various fields such as medicine, manufacturing, control, agriculture and academic applications. In recent decades, neural networks have been used to the identification, assessment and diagnosis of diseases. In this thesis we performed a comparative study among fuzzy neural networks (ANFIS), multilayer perceptron neural networks (MLP), radial basis function network (RBF) and generalized regression (GRNN) in the area of biomedical engineering and agronomy. In biomedical engineering neural networks and neuro fuzzy were trained and validated with data set from patients (91 subjects, 81 healthy and 10 hemiplegic). The GRNN network had the lowest Root Mean Square Error (RMSE), but the MLP network was able to identify a case of hemiplegia. In the area of agriculture a comparative study to estimate the wheat (Triticum aestivum) productivity was proposed using neural networks. For this study it was used data from an experimental database of wheat cultivars evaluated during two years in the region of Selvíria - MS. The validation was performed by comparing the estimated productivity through the quadratic regression curve and the output of the ANFIS with the neural networks. The RMSE error calculated with the GRNN and RBF neural networks was lower than that obtained with the quadratic regression and the ANFIS. The results obtained in the study of hemiplegia were validated using the RMSE, the confusion matrix, the sensitivity, the specificity and the error accuracy. The results showed that the use of neural networks and fuzzy neural networks, in biomedical engineering, can be a viable for monitoring the progress of patients and discovery new information through a combination of parameters. In agriculture this methodology can bring benefits in combining several evaluation parameters of production to optimize production while minimize financial costs in new plantations