This thread talks about the use of k-fold cross-validation for time series model selection where using the rolling basis approach is recommended, but I'm wondering how one would adapt this to a classification setting where we have a dataset with severe class imbalance?
Would one simply incorporate the stratified sampling approach with bootstrapping (to ensure that each fold is representative of all strata of the data), then apply the rolling basis technique?
Or are there other things one would need to consider for this situation?
