Holt’s Winters Exponential Smoothing (Additive and Multiplicative)
Holt-Winters is a seasonal forecasting procedure with more smoothing components than a Simple Exponential Smoothing model. While Simple Exponential Smoothing has one constant that smoothes both the trend and the seasonal components equally, the Holt-Winters model incorporates three potentially different smoothing constants. The Holt-Winters model has one smoothing constant to update the level, one for the slope, and one for the seasonal components. This approach is most useful when the seasonal component and the trend component are changing at different paces over time. The three smoothing constants lie somewhere between 0 and 1. If the components change rapidly, large smoothing constants should be considered; for stable components, the smoothing constants should be close to zero.