Simple Exponential Smoothing is a forecasting method that is not based on the analysis of the entire historical time series. Rather, Simple Exponential Smoothing uses a weighted moving average as the forecast, with the assigned weights decreasing exponentially for periods farther in the past. There are several types of exponential smoothing models. The following method describes Simple Exponential Smoothing.
a (alpha) is a smoothing constant, the most recent value of the time series is weighted by a , the next most recent value is weighted by a (1 – a ), the next value by a (1-a )^2, and so forth. All weighted values are then summed to determine the forecast. Typically, a is calculated by 1/(Number of actual data points), however a can be a chosen value. The closer the value of a is set to 1.0, the more heavily is the forecast weighted by the most recent results.
The formula for determining the forecast by the method of Simple Exponential Smoothing is: F(t+1) = F(t) + a [Y(t) ö F(t)] The definitions of the values used in simple exponential smoothing are as follows: F(t+1)= forecast for the next period F(t) = forecast for the most recent period a = smoothing constant (1 Ò a Ò 0) Y(t) = actual value for the most recent period. Simple Exponential Smoothing is most effective as a forecasting method when cyclical and irregular influences comprise the main effects on the time series values.