Box Jenkins model |

Two gentlemen named Box and Jenkins in 1976 developed this approach to ARIMA modeling. It is really not a forecasting model, but rather a procedure used to select from a group of forecasting models that best fit to the particular set of time series data. The strengths of the Box-Jenkins approach are its versatility - it can be used with most time series data - and its track record - its forecasting accuracy tends to exceed that of most time series models.

The Box-Jenkins approach involves three basic activities:

1) Identifying the tentative model

2) Determining the models parameters

3) Testing/applying the model

If the model developed in steps 2 and 3 does not meet expectations, the process is repeated and a new model is chosen and tested. The Box-Jenkins approach is more complicated than the other time series models, but this approach is also capable of handling almost any type of time series data. A number of studies comparing forecasting models indicate that the Box-Jenkins approach provides some of the more accurate short-range forecasts (one to three periods out) of any time series models.

The main features of the Box-Jenkins model are:

1) Its complexity discourages many forecasters and managers from using it.

2) It is best suited to short-range (i.e. daily, weekly or monthly) forecasts.

3) It requires a large amount of data (some authors feel at least sixty periods of data).

4) It is usually necessary to develop a new model whenever new sales data appear.