search

sillasgonzaga / mafs

Multiple Automatic Forecast Selection
GNU General Public License v2.0
19 stars 5 forks source link

Feature: Cross Validation #6

Open joshualeond opened 6 years ago

joshualeond commented 6 years ago

Thanks for open sourcing your work! I just had a comment on the model performance evaluation.

It looks like this package checks the best out-of-sample forecast for the last set of values of test_size. If that's the case then it could be beneficial to include a rolling forecast origin evaluation. Given that the forecast package has the tsCV() function you might be able to use that easily.

I'm sure this would make the model selection process much slower but would probably be more robust. It could just be an option in the select_forecast function.

sillasgonzaga commented 6 years ago

Thank you for using this package.

Cross Validation is indeed a very useful resource for forecasting evaluation, so I'll take a look into that, although I'm quite short on time. Please feel free to submit a pull request if you would to be a contributor for the package.