Just as double exponential smoothing expanded on simple exponential smoothing, triple exponential smoothing (also commonly referred to as Holt–Winters forecasting) builds on and expands the capabilities of Holt's linear method. This method is finally able to cope with seasonal data. It is called triple exponential smoothing because, for series with seasonality, there are now three signals to be removed (error, trend, and seasonality). Holt–Winters achieves this by adding yet another parameter, gamma, that controls the smoothing of the seasonal component. We will be letting the hw function choose the appropriate gamma value for us.
We will be performing Holt–Winters forecasting on both of our seasonal series examples. ...
