FORECASTING PERFORMANCE OF THE (MA) MODEL AND THE (GMA) MODEL WITH APPLICATIONS TO FINANCE

摘要

A new class of models known as the Generalized Moving Average (GMA) Model has been introduced in a previous paper by Peiris, et.al(2004) in order to reveal some hidden features, such as slowly changing local variances in time series data. The main objective of this study is to compare the forecasting performance of the Generalized Moving Average (GMA) model and the usual Moving Average (MA) model. We first compare the forecast of the data set of weekly sales of absorbent paper towels as reported in Bowerman and O'Connell (1987) using the standard MA against GMA model. Subsequently we compare the forecasting performance between MA and GMA models of two financial data sets. Our results indicate that the GMA(1) model has a better forecasting performance than the MA(1) model.