Week 4: The forecaster’s toolbox
What you will learn this week
- Forecasting workflow
- Four benchmark forecasting methods that we will use for comparison
- Linear trends, dummy seasonality
- Fitted values, residuals
- Forecasting using transformations
- Forecasting with decompositions
- Forecast evaluation
Pre-class activities
- Read Chapter 5 of the textbook and watch all embedded videos
- Read Chapter 7.4 of the textbook and watch the embedded video. Ignore Fourier series for the moment.
Slides for seminar
Workshop activities
Seminar Code
Tutorial exercises (in class or on your own)
NoteTutorial Learning Objectives
- Apply and compare simple benchmark forecasting methods.
- Able to select and implement appropriate baseline forecasting methods (e.g., NAIVE, SNAIVE, random walk with drift) for a variety of real-world time series and produce forecast plots.
- Evaluate forecasting performance using simple benchmark forecasting methods.
- Produce and visualise forecasts and evaluate whether a simple benchmark provides a reasonable baseline for forecasting.
- Decompose and forecast with STL methods.
- Apply STL decomposition to extract time series components.
- Able to generate seasonally adjusted forecasts, reseasonalise results, and compare performance with simple benchmark methods while interpreting residuals.
Complete Exercises 1, 3, 5, 11 from Section 5.11 of the book.
Complete Exercise 2 from Section 7.10 of the book.