Discussion on "Time Series Forecasting Lab (Part 6) - Stacked Ensembles"

Boris Guarisma · 2022-01-17T22:25:46.214Z

Cover photo by Markus Spiske on Unsplash Go to R-bloggers for R news and tutorials contributed by hundreds of R bloggers. Introduction This is the sixth of a series of 6 articles about time series forecasting with panel data and ensemble stacking wit...

Hi Boris Guarisma, thank you for a great tutorial, this is the most comprehensive guide to TS forecasting in R that I've come across. Is it possible for this code to be run on a spark cluster rather than locally? Thanks again.

Hi, forecast_stacking_tbl is also missing.

Boris Guarisma but it is used in lapply before the bind_rows.

Simona Jokubauskaite yes, my bad, I've skipped an important part, the article has been updated. It was my last article ... I was getting tired. Thanks !

Boris Guarisma Thanks for you great work! I really appreciate it :)

Hi Boris, From where comes submodels_resamples_kfold_tbl in 'Time Series Forecasting Lab (Part 6) - Stacked Ensembles' ? I'm currently using time_series_cv(), which seems to be wrong. Could you provide the complete script in cronological order e.g. on Github? https://github.com/guzu92/courses

Hi John, thank you for your message. Indeed, I forgot to explain how the resampled predictions are obtained, I have inserted a new code snippet above, see "Generate predictons for meta-learners". Use function modeltime_fit_resamples() which uses the k-fold resamples from vfold_cv().

Boris Guarisma Hello, this is different from Matt's documentation pages. I never attended his labs and have to compare/test your code against my self-thought construction. Till now, I get different .resample_results objects per model and therefor not all predictions for the metalearner. Thank you for your writings.

John Rambo Thanks ! FYI the article has been updated, mostly at the end (final forecast) where I've skipped an important part.

Search Hashnode

Time Series Forecasting Lab (Part 6) - Stacked Ensembles

Responses3