pmdarima.model_selection.SlidingWindowForecastCV

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

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

[R90]

https://robjhyndman.com/hyndsight/tscv/

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_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)

Initialize self. See help(type(self)) for accurate signature.

get_params(deep=True)

Get parameters for this estimator.

Parameters:

deep : bool, default=True

If True, will return the parameters for this estimator and contained subobjects that are estimators.

Returns:

params : mapping of string to any

Parameter names mapped to their values.

horizon

The forecast horizon for the cross-validator

set_params(**params)

Set the parameters of this estimator.

The method works on simple estimators as well as on nested objects (such as pipelines). The latter have parameters of the form <component>__<parameter> so that it’s possible to update each component of a nested object.

Parameters:

**params : dict

Estimator parameters.

Returns:

self : object

Estimator instance.

split(y, X=None)

Generate indices to split data into training and test sets

Parameters:

y : array-like or iterable, shape=(n_samples,)

The time-series array.

X : array-like, shape=[n_obs, n_vars], optional (default=None)

An optional 2-d array of exogenous variables.

Yields:

train : np.ndarray

The training set indices for the split

test : np.ndarray

The test set indices for the split

Examples using pmdarima.model_selection.SlidingWindowForecastCV

Cross-validation predictions

Cross-validating your time series models