Aplicação de redes neurais artificiais e Box-Jenkins na previsão de preços do feijão pagos ao produtor do estado de São Paulo : um estudo comparativo entre duas metodologias

Abstract

The studies of supply chains in Brazil are becoming increasingly common. Especially by your needs, but also driven by the complexity of economic issues, increase their competitiveness. The productive chain of the beans is one of the worst features scientific papers, which of course increases the relevance of this work. This paper tried to make an application and comparison of two methods for forecasting prices, artificial neural networks and Box and Jenkins, using a time series of prices of beans for the State of São Paulo (bag of 60 kg). This work involves the identification of deterministic patterns in time series with good results and flexibility, aiming at the projection series. The results showed that both approaches projections showed little variation compared with the actual series, and with very similar curves, moreover, the results obtained by means of artificial neural networks were statistically better when compared with those obtained by Box and Jenkins.