Stacked Ensembles (Breaking Changes)
The process for creating stacked ensembles is split into 2 steps:
modeltime_fit_resamples()
to generate resampled predictionsensemble_model_spec()
to apply stacking using a model_spec
Note - modeltime_refit(stacked_ensemble)
is still one step, which is the best way to handle refitting since multiple stacked models may have different submodel compositions. An additional argument, resamples
can be provided to train stacked ensembles made with ensemble_model_spec()
.
modeltime.ensemble
.