• split data into folds, with training and test data