Pycaret time series cv and binary classification

Question

is it possible to first use cross validation with ExpandingWindowSplitter on a time series dataset and then fit the folds on classification binary models?

Answer 1

PyCaret support custom CV generator object compatible with scikit-learn so you can try sktime's ExpandingWindowSplitter and pass it to the setup step. You can find documentation about PyCaret's classification parameter from here

from sktime.forecasting.model_selection import ExpandingWindowSplitter
from pycaret.classification import *

splitter = ExpandingWindowSplitter(fh=[2, 4], initial_window=5, step_length=2)
exp_name = setup(data = data,  target = 'target', fold_strategy=splitter)