Predictive Analytics: Microsoft Excel, 2nd Edition

5 More Advanced Smoothing Models

IN THIS CHAPTER

Holt’s Linear Exponential Smoothing

Seasonal Models

Using Additive Holt-Winters Models

Formulas for the Holt-Winters Additive and Multiplicative Models

The Models Compared

Damped Trend Forecasts

Holt’s Linear Exponential Smoothing

One way to account for a trend when you’re smoothing a time series is to use differencing, as discussed in the prior chapter. Most often, first differencing is enough to remove the trend from the time series, but sometimes second or even third differencing is needed to produce a stationary series.

I used the phrase “remove the trend” in the prior sentence because that’s what differencing does, and more advanced approaches to forecasting, such as Box-Jenkins ARIMA models, ...