WindowSummarizer

class WindowSummarizer(lag_config=None, lag_feature=None, n_jobs=- 1, target_cols=None, truncate=None)

Transformer for extracting time series features.

The WindowSummarizer transforms input series to features based on a provided dictionary of window summarizer, window shifts and window lengths.

Parameters

n_jobsint, optional (default=-1)

The number of jobs to run in parallel for applying the window functions. -1 means using all processors.

target_cols: list of str, optional (default = None)

Specifies which columns in X to target for applying the window functions. None will target the first column

lag_feature: dict of str and list, optional (default = dict containing first lag)

Dictionary specifying as key the type of function to be used and as value the argument window. For all keys other than lag, the argument window is a length 2 list containing the integer lag, which specifies how far back in the past the window will start, and the integer window length, which will give the length of the window across which to apply the function. For ease of notation, for the key “lag”, only a single integer specifying the lag argument will be provided.

Please see blow a graphical representation of the logic using the following symbols:

z = time stamp that the window is summarized to. Part of the window if lag is between 0 and 1-window_length, otherwise not part of the window. * = (other) time stamps in the window which is summarized x = observations, past or future, not part of the window

The summarization function is applied to the window consisting of * and potentially z.

For window = [1, 3], we have a lag of 1 and window_length of 3 to target the three last days (exclusive z) that were observed. Summarization is done across windows like this: |-------------------------- | | x x x x x x x x * * * z x | |---------------------------|

For window = [0, 3], we have a lag of 0 and window_length of 3 to target the three last days (inclusive z) that were observed. Summarization is done across windows like this: |-------------------------- | | x x x x x x x x * * z x x | |---------------------------|

Special case ´lag´: Since lags are frequently used and window length is redundant, a special notation will be used for lags. You need to provide a list of lag values, and window_length is not available. So window = [1] will result in the first lag:

|-------------------------- | | x x x x x x x x x x * z x | |---------------------------|

And window = [1, 4] will result in the first and fourth lag:

|-------------------------- | | x x x x x x x * x x * z x | |---------------------------|

key: either custom function call (to be

provided by user) or str corresponding to native pandas window function: * “sum”, * “mean”, * “median”, * “std”, * “var”, * “kurt”, * “min”, * “max”, * “corr”, * “cov”, * “skew”, * “sem” See also: https://pandas.pydata.org/docs/reference/window.html.

The column generated will be named after the key provided, followed by the lag parameter and the window_length (if not a lag).

second value (window): list of integers

List containg lag and window_length parameters.

truncate: str, optional (default = None)

Defines how to deal with NAs that were created as a result of applying the functions in the lag_feature dict across windows that are longer than the remaining history of data. For example a lag config of [14, 7] cannot be fully applied for the first 20 observations of the targeted column. A lag_feature of [[8, 14], [1, 28]] cannot be correctly applied for the first 21 resp. 28 observations of the targeted column. Possible values to deal with those NAs:

• None

• “bfill”

None will keep the NAs generated, and would leave it for the user to choose an estimator that can correctly deal with observations with missing values, “bfill” will fill the NAs by carrying the first observation backwards.

Returns

X: pd.DataFrame

Contains all transformed columns as well as non-transformed columns. The raw inputs to transformed columns will be dropped.

self: reference to self

Attributes

truncate_startint

See section Parameters - truncate for a more detailed explanation of truncation as a result of applying windows of certain lengths across past observations. Truncate_start will give the maximum of observations that are filled with NAs across all arguments of the lag_feature when truncate is set to None.

Examples

>>> import pandas as pd
>>> from sktime.transformations.series.summarize import WindowSummarizer
>>> from sktime.datasets import load_airline, load_longley
>>> from sktime.forecasting.naive import NaiveForecaster
>>> from sktime.forecasting.base import ForecastingHorizon
>>> from sktime.forecasting.compose import ForecastingPipeline
>>> from sktime.forecasting.model_selection import temporal_train_test_split
>>> y = load_airline()
>>> kwargs = {
...     "lag_feature": {
...         "lag": [1],
...         "mean": [[1, 3], [3, 6]],
...         "std": [[1, 4]],
...     }
... }
>>> transformer = WindowSummarizer(**kwargs)
>>> y_transformed = transformer.fit_transform(y)

Example where we transform on a different, later test set:

>>> y = load_airline()
>>> y_train, y_test = temporal_train_test_split(y)
>>> kwargs = {
...     "lag_config": {
...         "lag": ["lag", [[1, 0]]],
...         "mean": ["mean", [[3, 0], [12, 0]]],
...         "std": ["std", [[4, 0]]],
...     }
... }
>>> transformer = WindowSummarizer(**kwargs)
>>> y_test_transformed = transformer.fit(y_train).transform(y_test)

Example with transforming multiple columns of exogeneous features

>>> y, X = load_longley()
>>> y_train, y_test, X_train, X_test = temporal_train_test_split(y, X)
>>> fh = ForecastingHorizon(X_test.index, is_relative=False)
>>> # Example transforming only X
>>> pipe = ForecastingPipeline(
...     steps=[
...         ("a", WindowSummarizer(n_jobs=1, target_cols=["POP", "GNPDEFL"])),
...         ("b", WindowSummarizer(n_jobs=1, target_cols=["GNP"], **kwargs)),
...         ("forecaster", NaiveForecaster(strategy="drift")),
...     ]
... )
>>> pipe_return = pipe.fit(y_train, X_train)
>>> y_pred1 = pipe_return.predict(fh=fh, X=X_test)

Example with transforming multiple columns of exogeneous features as well as the y column

>>> Z_train = pd.concat([X_train, y_train], axis=1)
>>> Z_test = pd.concat([X_test, y_test], axis=1)
>>> pipe = ForecastingPipeline(
...     steps=[
...         ("a", WindowSummarizer(n_jobs=1, target_cols=["POP", "TOTEMP"])),
...         ("b", WindowSummarizer(**kwargs, n_jobs=1, target_cols=["GNP"])),
...         ("forecaster", NaiveForecaster(strategy="drift")),
...     ]
... )
>>> pipe_return = pipe.fit(y_train, Z_train)
>>> y_pred2 = pipe_return.predict(fh=fh, X=Z_test)

Methods

check_is_fitted()	Check if the estimator has been fitted.
clone_tags(estimator[, tag_names])	clone/mirror tags from another estimator as dynamic override.
create_test_instance([parameter_set])	Construct Estimator instance if possible.
create_test_instances_and_names([parameter_set])	Create list of all test instances and a list of names for them.
fit(X[, y])	Fit transformer to X, optionally to y.
fit_transform(X[, y])	Fit to data, then transform it.
get_class_tag(tag_name[, tag_value_default])	Get tag value from estimator class (only class tags).
get_class_tags()	Get class tags from estimator class and all its parent classes.
get_params([deep])	Get parameters for this estimator.
get_tag(tag_name[, tag_value_default, …])	Get tag value from estimator class and dynamic tag overrides.
get_tags()	Get tags from estimator class and dynamic tag overrides.
get_test_params([parameter_set])	Return testing parameter settings for the estimator.
inverse_transform(X[, y])	Inverse transform X and return an inverse transformed version.
is_composite()	Check if the object is composite.
reset()	Reset the object to a clean post-init state.
set_params(**params)	Set the parameters of this estimator.
set_tags(**tag_dict)	Set dynamic tags to given values.
transform(X[, y])	Transform X and return a transformed version.
update(X[, y, update_params])	Update transformer with X, optionally y.

classmethod get_test_params(parameter_set='default')

Return testing parameter settings for the estimator.

Parameters

parameter_setstr, default=”default”

Name of the set of test parameters to return, for use in tests. If no special parameters are defined for a value, will return “default” set.

Returns

paramsdict or list of dict, default = {}

Parameters to create testing instances of the class Each dict are parameters to construct an “interesting” test instance, i.e., MyClass(**params) or MyClass(**params[i]) creates a valid test instance. create_test_instance uses the first (or only) dictionary in params

check_is_fitted()

Check if the estimator has been fitted.

Raises

NotFittedError

If the estimator has not been fitted yet.

clone_tags(estimator, tag_names=None)

clone/mirror tags from another estimator as dynamic override.

Parameters

estimatorestimator inheriting from :class:BaseEstimator

tag_namesstr or list of str, default = None

Names of tags to clone. If None then all tags in estimator are used as tag_names.

Returns

Self

Reference to self.

Notes

Changes object state by setting tag values in tag_set from estimator as dynamic tags in self.

classmethod create_test_instance(parameter_set='default')

Construct Estimator instance if possible.

Parameters

parameter_setstr, default=”default”

Name of the set of test parameters to return, for use in tests. If no special parameters are defined for a value, will return “default” set.

Returns

instanceinstance of the class with default parameters

Notes

get_test_params can return dict or list of dict. This function takes first or single dict that get_test_params returns, and constructs the object with that.

classmethod create_test_instances_and_names(parameter_set='default')

Create list of all test instances and a list of names for them.

Parameters

parameter_setstr, default=”default”

Name of the set of test parameters to return, for use in tests. If no special parameters are defined for a value, will return “default” set.

Returns

objslist of instances of cls

i-th instance is cls(**cls.get_test_params()[i])

nameslist of str, same length as objs

i-th element is name of i-th instance of obj in tests convention is {cls.__name__}-{i} if more than one instance otherwise {cls.__name__}

parameter_setstr, default=”default”

Name of the set of test parameters to return, for use in tests. If no special parameters are defined for a value, will return “default” set.

fit(X, y=None)

Fit transformer to X, optionally to y.

State change:

Changes state to “fitted”.

Writes to self:

Sets is_fitted flag to True. Sets fitted model attributes ending in “_”.

Parameters

XSeries or Panel, any supported mtype

Data to fit transform to, of python type as follows:

Series: pd.Series, pd.DataFrame, or np.ndarray (1D or 2D) Panel: pd.DataFrame with 2-level MultiIndex, list of pd.DataFrame,

nested pd.DataFrame, or pd.DataFrame in long/wide format

subject to sktime mtype format specifications, for further details see

examples/AA_datatypes_and_datasets.ipynb

ySeries or Panel, default=None

Additional data, e.g., labels for transformation

Returns

selfa fitted instance of the estimator

fit_transform(X, y=None)

Fit to data, then transform it.

Fits the transformer to X and y and returns a transformed version of X.

State change:

Changes state to “fitted”.

Writes to self:

Sets is_fitted flag to True. Sets fitted model attributes ending in “_”.

Parameters

XSeries or Panel, any supported mtype

Data to be transformed, of python type as follows:

Series: pd.Series, pd.DataFrame, or np.ndarray (1D or 2D) Panel: pd.DataFrame with 2-level MultiIndex, list of pd.DataFrame,

nested pd.DataFrame, or pd.DataFrame in long/wide format

subject to sktime mtype format specifications, for further details see

examples/AA_datatypes_and_datasets.ipynb

ySeries or Panel, default=None

Additional data, e.g., labels for transformation

Returns

transformed version of X

type depends on type of X and scitype:transform-output tag:

X | tf-output | type of return |

|----------|————–|------------------------| | Series | Primitives | pd.DataFrame (1-row) | | Panel | Primitives | pd.DataFrame | | Series | Series | Series | | Panel | Series | Panel | | Series | Panel | Panel |

instances in return correspond to instances in X

combinations not in the table are currently not supported

Explicitly, with examples:

if X is Series (e.g., pd.DataFrame) and transform-output is Series

then the return is a single Series of the same mtype Example: detrending a single series

if X is Panel (e.g., pd-multiindex) and transform-output is Series

then the return is Panel with same number of instances as X

(the transformer is applied to each input Series instance)

Example: all series in the panel are detrended individually

if X is Series or Panel and transform-output is Primitives

then the return is pd.DataFrame with as many rows as instances in X Example: i-th row of the return has mean and variance of the i-th series

if X is Series and transform-output is Panel

then the return is a Panel object of type pd-multiindex Example: i-th instance of the output is the i-th window running over X

classmethod get_class_tag(tag_name, tag_value_default=None)

Get tag value from estimator class (only class tags).

Parameters

tag_namestr

Name of tag value.

tag_value_defaultany type

Default/fallback value if tag is not found.

Returns

tag_value

Value of the tag_name tag in self. If not found, returns tag_value_default.

classmethod get_class_tags()

Get class tags from estimator class and all its parent classes.

Returns

collected_tagsdict

Dictionary of tag name : tag value pairs. Collected from _tags class attribute via nested inheritance. NOT overridden by dynamic tags set by set_tags or mirror_tags.

get_params(deep=True)

Get parameters for this estimator.

Parameters

deepbool, default=True

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

Returns

paramsdict

Parameter names mapped to their values.

get_tag(tag_name, tag_value_default=None, raise_error=True)

Get tag value from estimator class and dynamic tag overrides.

Parameters

tag_namestr

Name of tag to be retrieved

tag_value_defaultany type, optional; default=None

Default/fallback value if tag is not found

raise_errorbool

whether a ValueError is raised when the tag is not found

Returns

tag_value

Value of the tag_name tag in self. If not found, returns an error if raise_error is True, otherwise it returns tag_value_default.

Raises

ValueError if raise_error is True i.e. if tag_name is not in self.get_tags(

).keys()

get_tags()

Get tags from estimator class and dynamic tag overrides.

Returns

collected_tagsdict

Dictionary of tag name : tag value pairs. Collected from _tags class attribute via nested inheritance and then any overrides and new tags from _tags_dynamic object attribute.

inverse_transform(X, y=None)

Inverse transform X and return an inverse transformed version.

Currently it is assumed that only transformers with tags

“scitype:transform-input”=”Series”, “scitype:transform-output”=”Series”,

have an inverse_transform.

State required:

Requires state to be “fitted”.

Accesses in self:

Fitted model attributes ending in “_”. self._is_fitted

Parameters

XSeries or Panel, any supported mtype

Data to be inverse transformed, of python type as follows:

Series: pd.Series, pd.DataFrame, or np.ndarray (1D or 2D) Panel: pd.DataFrame with 2-level MultiIndex, list of pd.DataFrame,

nested pd.DataFrame, or pd.DataFrame in long/wide format

subject to sktime mtype format specifications, for further details see

examples/AA_datatypes_and_datasets.ipynb

ySeries or Panel, default=None

Additional data, e.g., labels for transformation

Returns

inverse transformed version of X

of the same type as X, and conforming to mtype format specifications

is_composite()

Check if the object is composite.

A composite object is an object which contains objects, as parameters. Called on an instance, since this may differ by instance.

Returns

composite: bool, whether self contains a parameter which is BaseObject

property is_fitted

Whether fit has been called.

reset()

Reset the object to a clean post-init state.

Equivalent to sklearn.clone but overwrites self. After self.reset() call, self is equal in value to type(self)(**self.get_params(deep=False))

Detail behaviour: removes any object attributes, except:

hyper-parameters = arguments of __init__ object attributes containing double-underscores, i.e., the string “__”

runs __init__ with current values of hyper-parameters (result of get_params)

Not affected by the reset are: object attributes containing double-underscores class and object methods, class attributes

set_params(**params)

Set the parameters of this estimator.

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

Parameters

**paramsdict

Estimator parameters.

Returns

selfestimator instance

Estimator instance.

set_tags(**tag_dict)

Set dynamic tags to given values.

Parameters

tag_dictdict

Dictionary of tag name : tag value pairs.

Returns

Self

Reference to self.

Notes

Changes object state by settting tag values in tag_dict as dynamic tags in self.

transform(X, y=None)

Transform X and return a transformed version.

State required:

Requires state to be “fitted”.

Accesses in self:

Fitted model attributes ending in “_”. self._is_fitted

Parameters

XSeries or Panel, any supported mtype

Data to be transformed, of python type as follows:

Series: pd.Series, pd.DataFrame, or np.ndarray (1D or 2D) Panel: pd.DataFrame with 2-level MultiIndex, list of pd.DataFrame,

nested pd.DataFrame, or pd.DataFrame in long/wide format

subject to sktime mtype format specifications, for further details see

examples/AA_datatypes_and_datasets.ipynb

ySeries or Panel, default=None

Additional data, e.g., labels for transformation

Returns

transformed version of X

type depends on type of X and scitype:transform-output tag:

| transform | |

X | -output | type of return |

|----------|————–|------------------------| | Series | Primitives | pd.DataFrame (1-row) | | Panel | Primitives | pd.DataFrame | | Series | Series | Series | | Panel | Series | Panel | | Series | Panel | Panel |

instances in return correspond to instances in X

combinations not in the table are currently not supported

Explicitly, with examples:

if X is Series (e.g., pd.DataFrame) and transform-output is Series

then the return is a single Series of the same mtype Example: detrending a single series

if X is Panel (e.g., pd-multiindex) and transform-output is Series

then the return is Panel with same number of instances as X

(the transformer is applied to each input Series instance)

Example: all series in the panel are detrended individually

if X is Series or Panel and transform-output is Primitives

then the return is pd.DataFrame with as many rows as instances in X Example: i-th row of the return has mean and variance of the i-th series

if X is Series and transform-output is Panel

then the return is a Panel object of type pd-multiindex Example: i-th instance of the output is the i-th window running over X

update(X, y=None, update_params=True)

Update transformer with X, optionally y.

State required:

Requires state to be “fitted”.

Accesses in self:

Fitted model attributes ending in “_”. self._is_fitted

Writes to self:

May update fitted model attributes ending in “_”.

Parameters

XSeries or Panel, any supported mtype

Data to fit transform to, of python type as follows:

Series: pd.Series, pd.DataFrame, or np.ndarray (1D or 2D) Panel: pd.DataFrame with 2-level MultiIndex, list of pd.DataFrame,

nested pd.DataFrame, or pd.DataFrame in long/wide format

subject to sktime mtype format specifications, for further details see

examples/AA_datatypes_and_datasets.ipynb

ySeries or Panel, default=None

Additional data, e.g., labels for transformation

update_paramsbool, default=True

whether the model is updated. Yes if true, if false, simply skips call. argument exists for compatibility with forecasting module.

Returns

selfa fitted instance of the estimator