Autoregression (Seasonal)

The autoregression forecasting method is based on the auto-correlational approach to time series forecasting. Autoregression forecasting detects the linear, nonlinear, and seasonal fluctuations in historic data and projects these trends into the future. Autoregression provides the best forecasting reliability when the driving factors underlying your business are affected by seasonal fluctuations.

A multiline plot of time and revenue will show up-and-down fluctuations that may reflect seasonal variations. For example, if your revenues are growing exponentially due to the introduction of a best selling product, but sales of that product are also seasonal, then autoregression forecasting provides a more reliable forecast than the Growth method.

Use the autoregression method when you have historic data representing a large number of time periods (for example, more than 24 monthly periods) and when seasonal variations may occur in it.

For crosstabs, if you nest multiple levels of time, IBM® Cognos® PowerPlay® produces the forecast only at the highest level of time. For example, if you nest quarters within years for revenue and then insert a forecast calculation, PowerPlay generates the forecast only at the years level. To generate your forecast at the quarters level, delete the years level before you generate the forecast.

If you applied ranking in your crosstab, PowerPlay creates the forecast you request, however, forecasts are not included in the rankings.

If you convert the currency in your crosstab, PowerPlay creates the forecast on the currency-converted values.