In this section, I shall explain a modified form of the Holt-Winters exponential smoothing algorithm, which is quite useful for forecasting. Holt-Winters is a fairly simple algorithm. Here it is:
func hw(a stl.Result, periodicity, forward int, alpha, beta, gamma float64) []float64 { level := make([]float64, len(a.Data)) trend := make([]float64, len(a.Trend)) seasonal := make([]float64, len(a.Seasonal)) forecast := make([]float64, len(a.Data)+forward) copy(seasonal, a.Seasonal) for i := range a.Data { if i == 0 { continue } level[i] = alpha*a.Data[i] + (1-alpha)*(level[i-1]+trend[i-1]) trend[i] = beta*(level[i]-level[i-1]) + (1-beta)*(trend[i-1]) if i-periodicity < 0 { continue } seasonal[i] = gamma*(a.Data[i]-level[i-1]-trend[i-1]) ...
