Box-Jenkins models

Definition

Mathematical models used typically for accurate short-term forecasts of 'well-behaved' data (that shows predictable repetitive cycles and patterns). B-J models require at least a moderately long time series (with about a hundred observations) for an effective 'fitting,' and generally include autoregressive, moving average, and seasonal moving average terms, and difference and seasonal difference operators. Both ARMA model and ARIMA model are commonly called Box-Jenkins models after the US mathematicians George Box and Gwilym Jenkins who popularized them in their 1976 book 'Time Series Analysis-Forecasting And Control.'