pmdarima.model_selection.SlidingWindowForecastCV

class pmdarima.model_selection.SlidingWindowForecastCV(h=1, step=1, window_size=None)[source][source]

Use a sliding window to perform cross validation

This approach to CV slides a window over the training samples while using several future samples as a test set. While similar to the RollingForecastCV, it differs in that the train set does not grow, but rather shifts.

Parameters:

h : int, optional (default=1)

The forecasting horizon, or the number of steps into the future after the last training sample for the test set.

step : int, optional (default=1)

The size of step taken between training folds.

window_size : int or None, optional (default=None)

The size of the rolling window to use. If None, a rolling window of size n_samples // 5 will be used.

Attributes

horizon

The forecast horizon for the cross-validator

See also

RollingForecastCV

References

Examples

With a step size of one and a forecasting horizon of one, the training size will grow by 1 for each step, and the test index will be 1 + the last training index. Notice the sliding window also adjusts where the training sample begins for each fold:

>>> import pmdarima as pm
>>> from pmdarima.model_selection import SlidingWindowForecastCV
>>> wineind = pm.datasets.load_wineind()
>>> cv = SlidingWindowForecastCV()
>>> cv_generator = cv.split(wineind)
>>> next(cv_generator)
(array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,
       17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,
       34]), array([35]))
>>> next(cv_generator)
(array([ 1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17,
       18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34,
       35]), array([36]))

With a step size of 4, a forecasting horizon of 6, and a window size of 12, the training size will grow by 4 for each step, and the test index will 6 + the last index in the training fold:

>>> cv = SlidingWindowForecastCV(step=4, h=6, window_size=12)
>>> cv_generator = cv.split(wineind)
>>> next(cv_generator)
(array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11]),
 array([12, 13, 14, 15, 16, 17]))
>>> next(cv_generator)
(array([ 4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15]),
 array([16, 17, 18, 19, 20, 21]))

Methods

get_metadata_routing()

Get metadata routing of this object.

get_params([deep])

Get parameters for this estimator.

set_params(**params)

Set the parameters of this estimator.

split(y[, X])

Generate indices to split data into training and test sets
__init__(h=1, step=1, window_size=None)[source][source]

Examples using pmdarima.model_selection.SlidingWindowForecastCV

Cross-validation predictions

Cross-validation predictions

Cross-validating your time series models

Cross-validating your time series models