The Decomposition model is used to identify underlying components by breaking the series into its component parts (trend, cycle, seasonal variations and irregular fluctuations) and then reassembling the parts to construct a forecast.
Trend is the long-term component representing the growth or decline in the time series over an extended period of time (i.e. price inflation or population growth). The cyclical component is an upward and downward change, occurring over a period of two to 10 years or longer (i.e. economic expansions and contractions). The seasonal component is repeated annually, and reflects weather, holidays and/or length of the calendar year (like the seasonal fluxes of retail sales). After the other components have been removed, the irregular component measures the variability of the time series based on unpredictable factors like major weather changes or strikes. In other words, the irregular component accounts for randomness. The goal of decomposition is to extract and take into account each influential factor in a time series and obtain a forecast for each. Then, an overall forecast is achieved by combining the projections for each component. So, rather than having only one defined measurement (time), decomposition justifies X’s value over time, forecasting from specific components within time.
To use the Decomposition forecasting technique:
- Click on the Forecast Method tab.
- In the Forecast Technique area, scroll through the list of methods and select Decomposition. The Decomposition Forecasting technique displays.
- Select Edit parameters to activate Decomposition’s parameters. The following table details what each parameter means.
Parameter Description Type
Indicate the type of Seasonality of the Decomposition method.
- Multiplicative - Seasonality is Multiplicative.- Additive - Seasonality is Additive.
Forecast Method for Decomposed Data Select the method used to forecast the long-term trend and cyclical movement in the decomposed data.
- Click Finish.