Ondaietas e previsão de séries de tempo: uma análise empírica

Abstract

This paper presents three case studies in time series forecasting. We try to compare the use of traditional ARIMA models with an alternative method that combines of ARIMA and Wavelets models. Two different approaches are applied. In the first one, Wavelets are used to fraction the original time series, so that ARIMA forecasting is performed on the ffactioned series. The fractioned forecasting is then jointed to obtain the original series forecasting. The second alternative method consist in using Wavelets to smooth the original series before using traditional ARIMA forecasting.