Transfer function models are widely used in engineering and in economics. In this article, an automatic procedure is proposed to identify such models. The proposed procedure can directly handle nonstationarity, outliers and other deterministic effects such as Trading Day or Easter. The procedure is applied to one simulated series and one real series.

Keywords: ARIMA model, Forecasting, Information Criterion, Transfer Function.

AMS Subject classifications: 62M10, 91B84, 62B10, 94A17.