Previsão de séries de tempo: redes neurais artificiais e modelos estruturais

Abstract

This article has a dual objective: on one hand we present the Artifitial Neural Network (ANN) model from a statistical perspective; an the other hand, we also present two forecasting exercises to alow an empirical comparison between the ANN model and the traditional structural time series or unobservable components model (UCM). We use the UCM as a way to help determining the topology of the ANN. The results do not show clear cut conclusion as far as one step ahead forecasts are concerned, but for longer time horizons the ANN has a better forecasting performance.