Hi! If we use Conformal Prediction (e.g.,

ConformalIntervals(h=24, n_windows=2)

) for uncertainty estimation, is it still valid to perform cross-validation for model selection? Does conformal prediction already incorporate a validation step that overlaps with cross-validation?

It doesn't overlap, it's always based on the past

Thanks for your response. Since conformal prediction is always based on the past, how can we structure cross-validation while also using conformal intervals, specifically in a hierarchical context? Do you have any resources on this?