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 crossvalidator See also
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)[source][source]¶ Initialize self. See help(type(self)) for accurate signature.

get_params
(deep=True)[source]¶ 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 crossvalidator

set_params
(**params)[source]¶ 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)[source]¶ Generate indices to split data into training and test sets
Parameters: y : arraylike or iterable, shape=(n_samples,)
The timeseries array.
X : arraylike, shape=[n_obs, n_vars], optional (default=None)
An optional 2d 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
