Walk forward cross-validation in tsai

[alitirmizi23]

I was wondering if there are plans to implement walk forward CV in TSAI. #

[vrodriguezf]

By default, the function get_splits with n_splits > 0 creates disjunct splits, so ou can use them as folds in a CV process.

[oguiza]

Hi @alitirmizi23, cc: @vrodriguezf ,
Thanks for creating this issue.
This is something I've had in mind for quite some time. I've found some time this morning to add this new functionality. You can use it by doing this:

from tsai.data.validation import get_walk_forward_splits
splits = get_walk_forward_splits(
    o, 
    n_splits=4, 
    train_size=.6,
    valid_size=0.1, 
    test_size=0.1, 
    anchored = True,
    gap = 0.,
    test_after_valid = True,
    random_state = None,
    show_plot=True,
)

You can find the documentation here.

[oguiza]

@alitirmizi23 , if this is what you were looking for, pls feel free to close the issue.

[alitirmizi23]

@oguiza amazing, thanks for the feature! Yes, this is what I was looking for. I will close this now.