Redes neurais artificiais aplicadas à previsão da incidência de malária no estado de Roraima

• Ano de defesa: 2010
• Autor(a) principal: Cunha, Guilherme Bernardino da
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; BR; Programa de Pós-graduação em Engenharia Elétrica; Engenharias; UFU
• Palavras-chave em Português: Redes neurais artificiais; Previsão da incidência de malária; Backpropagation; Modelo ARIMA; Suavização exponencial; Redes neurais (Computação); Malária - Roraima; Artificial neural network; Forecasting of malaria; ARIMA models; Exponential smoothing; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: https://repositorio.ufu.br/handle/123456789/14275

The present work aims to create a prototype called SISPIMA - forecast system in the incidence of malaria, to generate estimates of the incidence of malaria in Roraima state in three different periods: short term (3 months), medium term (6 months) and long term (12 months). To develop the system, were employed techniques of artificial neural networks and time series analysis. The SISPIMA consists of four steps: collection and storage of data, preprocessing, training and predicting the incidence of malaria. Data were obtained through access to the site SIVEP-Malaria Health Ministry. These were filtered, normalized and classified by SISPIMA in the pre-processing before performing the training and prediction. For training and forecasting, used artificial neural networks. The architecture of artificial neural network used was the multilayer perceptron (MLP) with a variation of the backpropagation training algorithm, called of Resilient Propagation (RPROG). To validate the results and assess the performance and accuracy of the proposed system, we use the ARIMA model as a comparison because of its wide application in epidemiological time series forecasting.