Recursive Models for Forecasting Seasonal Processes

Abstract

Many of the typical problems encountered in forecasting a time series are alleviated when the series follows a seasonal pattern. The seasonal effect is independent of a long-term trend and cyclic effect. Furthermore, since the seasonal effect is recurrent and periodic, it is predictable. Thus, when the time series follows a seasonal pattern, the general shape of the series is known. Questions regarding the growth trend of the series, the expansion and contraction of the seasonal pattern, and random variation affecting the series are, however, left unanswered.