The final model is trained on the sampled dataset. The metrics used to rank models obtained by different algorithms are computed on the test set.

The metrics used to rank models obtained by different algorithms are computed on each of the test folds. The final model is trained on the sampled dataset.

The offset between consecutive cross-test evaluation folds aims at avoiding overlaps with hyperparameter search validation fold that could cause an optimistic bias on model evaluation metrics.

Every train set fold is of equal duration.

The metrics used to rank hyperparameter points are computed by simple split validation. This strategy holds out the last forecast horizon time steps and tunes models on the validation set.

The metrics used to rank hyperparameter points are computed by cross-validation on each of the validation folds.

The offset between consecutive hyperparameter search validation folds aims at avoiding overlaps with cross-test evaluation folds that could cause an optimistic bias on model evaluation metrics.

Every train set fold is of equal duration.