quantlab.model.walk_forward_training.train_walk_forward

quantlab.model.walk_forward_training.train_walk_forward(unit, train_periods, expanding=False, test_periods=None)

Run a walk-forward cross-validation of unit and return the walk-forward run.

See the module docstring for the procedure; the parameters are those of walk_forward_folds.

Parameters:

unitWalkForwardTrainable

The collected model or ensemble.

train_periodsint

Number of bars in the first fold’s training window, and in every fold’s when sliding.

expandingbool, default False

Train every fold from the first fold’s start.

test_periodsint, optional

Number of bars in each fold’s test window, and the step between folds. Defaults to train_periods // 5.

Returns:

TrainedRun

The "walk_forward" run: its folds and cv_mean.

Raises:

ValueError

If a hyperparameter is invalid, walk_forward_folds refuses the settings (checked first), or the unit has no bar in its date range; nothing is created on disk then.

Examples

>>> run = train_walk_forward(model.collect(), train_periods=20, test_periods=10)
>>> len(run.folds), sorted(run.cv_mean)[:1]
(2, ['cv_mean_test_ic'])