Feature Selection with BorutaShap

[mdo]

@adalseno

I mean, even when one does Cross Validation (and Boruta-Shap does not), one splits the series in chunks and for every chunk will use the first part for training and the last part for testing.

This is not true for standard CV, but is for walk forward CV. A era-wise walk forward/time series split could also be used here as you point out @objectscience and could be worth checking out. You could probably use the splitter from the walk forward/time series CV post.