JSzitas / soothsayer

Automatic Time Series Forecasting and Ensembling via Meta-learning
Other
1 stars 0 forks source link

`add_soothsayer` does not work correctly with a keyed mable and multiple time series #5

Closed JSzitas closed 2 years ago

JSzitas commented 2 years ago

The first line

.data <- mable[[1]][[1]][["data"]] is almost surely the problem.

We can get model names as fabletools::mable_vars(mable), group_by(key) and recover data that way. thus our result will be a dataset which is a reconstruction of the original dataset.

We might face further issues with iteration over multiple data, and might need to replace e.g. bind_cols() with a _join()

JSzitas commented 2 years ago

Should be resolved by 3b6146a