niadel91/forecasting_tourism_in_australia

Analysis of different Forecasting techniques on a time series dataset to forecast the number of tourists in Australia in R

Forecasting_Tourism_in_Australia

Analysis of different Forecasting techniques on a time series dataset to forecast the number of tourists in Australia

Data Source

visitors dataset from fpp2 package. These data include monthly Australian shortterm overseas visitors data, May 1985–April 2005.

Approach

1. Made a Time Plot of the data, describing the main features of the series.
2. Split the data into a training set and a test set comprising the last two years of available data.
3. Forecast the test set using Holt-Winters’ multiplicative method and used the output to analyse whether Multiplicative Seasonality is necessary here.
4. Forecast the two-year test set using each of the following methods:
   • ETS Model
   • Additive ETS model applied to a Box-Cox transformed series
   • Seasonal naïve method
5. Which method gives the best forecasts? Does this method pass the residuals test?

Conclusion

Surprisingly, the Seasonal naïve method was able to provide the best forecasts but it could not pass the residuals test.