Intro to Forecasting - Part 3 - HRUG
- 1. Intro to Forecasting in R Part 3 Houston R Users Group Ed Goodwin, CFA
- 2. The story so far… • time series classes in R • simple and naive forecasts • simple linear regressions • dummy variables • data transformations
- 3. One more transformation using scale( ) Use the scale function to scale two different matrices. Scale uses the formula (x-mean(x))/ sd(x) to make all data sets on same scale. You can use arguments to set the whether you center on 0 or not and whether to scale by a certain numeric amount or default to standard deviation
- 4. Example of Scale in Action • Monthly time series of pigs slaughtered • Monthly time series of pounds of milk produced
- 6. Scaling code
- 7. Scaled plot
- 8. Stationarity Stationarity is a measure of how much a point in a time series is dependent on prior points in the time series. You must know if your data is stationary (or non-stationary) in order to determine how to forecast. Certain models only work on stationary time series (e.g. ARIMA)
- 9. Stationarity - Examples Non-Stationary Stationary
- 10. Use the acf( ) function Use acf( ) to generate an auto correlation plot to analyze stationarity
- 11. Use the acf( ) function Use acf( ) to generate an auto correlation plot to analyze stationarity
- 12. Decomposing the AirPassengers Data Series
- 13. It’s easy…just use the decompose( ) function type can be additive or multiplicative
- 14. Exponential Smoothing Using Holt-Winters Holt-Winters is a seasonal trend forecasting model. It is a simple exponential smoothing model. This means that it predicts the future based on a set of smoothed average observations about the past, accounting for trend and seasonal variations.
- 15. Holt-Winters Use the hw( ) function.
- 16. Holt-Winters Forecast of Air Passengers