AR model prediction of time series with trends and seasonalities: A contrast with Box-Jenkins modeling

A "long autoregressive (AR) model alternative to the classical Box-Jenkins ARIMA model method of modeling time series with trend and seasonality characteristics is considered. Superior forecast performance is demonstrated by our long AR model method on the Box-Jenkins Series G airline passenger data. The difference in performance is accounted for by the relative underparameterization of the Box-Jenkins method. A Householder transformation-Akaike AIC criterion method is employed for determining the best data transformed, detrended-deseasonalized stationary residuals-AR modeled time series.