Model specification and forecasting accuracy of some seasonal interaction models: an empirical investigation using the "airiline" data

Abstract

Two well know approaches to seasonal forecasting, are the Holt-Winters and Box-Jenkins methods. Forecasts obtained by these methods, are com¬pared with the forecasts from some seasonal interaction models (transfer-function type ARIMA models employing trend and seasonal dummy variables). The advantages of each method can be judged from their goodness of fit and their ex-post forecasting accuracy in the particular case of the "airline" data.