library(fpp3)
<- tourism |>
aus_trips summarise(Trips = sum(Trips))
Activities: Week 7
This week we will forecast total domestic overnight trips across Australia from the tourism
dataset.
- Plot the data and describe the main features of the series.
- Use
ETS()
to choose a seasonal model for the data. What do the parameters tell you about the series? - Forecast the next two years using your ETS model.
- Decompose the series using STL and obtain the seasonally adjusted data.
- Forecast the next two years of the series using an ETS(A,A,N) model applied to the seasonally adjusted data (specified using
decomposition_model()
.) - Which of the two models gives the best forecasts? Which gives the best fit to the training data?
- Now set up a test set of 3 years and fit both models to the training set, along with at least one appropriate benchmark.
- Compute the forecast accuracy of the three models. Which is best?
- Check the residuals of your preferred model using
gg_tsresiduals()
and Ljung-Box test.