Week 6: Exponential smoothing
What you will learn this week
- Exponential smoothing methods with trend and seasonality
- ETS models
- Automatic model selection using the AICc
Pre-seminar activities
Read Sections 8.3-8.7 of the textbook and watch all embedded videos.
Tutorial exercises
NoteTutorial Learning Objectives
- Compare alternative ETS model structures.
- Select and fit different ETS models (e.g., ETS(A,N,N) vs ETS(A,A,N)).
- Evaluate and compare their fit (e.g., RMSE).
- Interpret forecast differences and prediction intervals, and justify model choice based on empirical performance.
- Select and justify appropriate ETS models for seasonal data.
- Able to identify when multiplicative seasonality (and optional damped trend) is necessary.
- Compare multiple forecasting methods using training/test splits and cross-validation.
- Use training/test sets and cross-validation to fit and compare a range of methods (ETS, transformed ETS, seasonal naïve, STL + ETS).
- Evaluate ETS method and identify its limitations.
- Apply ETS model, assess forecasting quality, and recognise cases where ETS may not perform well and explain why.
- Complete Exercises 5, 7, 11, 14 from Section 8.8 of the book.