_OneSegmentChangePointsTrendTransform

class _OneSegmentChangePointsTrendTransform(in_column: str, change_point_model: ruptures.base.BaseEstimator, detrend_model: Type[sklearn.base.RegressorMixin], **change_point_model_predict_params)

Bases: etna.transforms.base.Transform

_OneSegmentChangePointsTransform subtracts multiple linear trend from series.

Init _OneSegmentChangePointsTrendTransform.

Parameters

• in_column (str) – name of column to apply transform to

• change_point_model (ruptures.base.BaseEstimator) – model to get trend change points TODO: replace this parameters with the instance of BaseChangePointsModelAdapter in ETNA 2.0

• detrend_model (Type[sklearn.base.RegressorMixin]) – model to get trend in data

• change_point_model_predict_params – params for change_point_model.predict method

Inherited-members

Methods

fit(df)	Fit OneSegmentChangePointsTransform: find trend change points in df, fit detrend models with data from intervals of stable trend.
fit_transform(df)	May be reimplemented.
inverse_transform(df)	Split df to intervals of stable trend according to previous change point detection and add trend to each one.
load(path)	Load an object.
save(path)	Save the object.
to_dict()	Collect all information about etna object in dict.
transform(df)	Split df to intervals of stable trend and subtract trend from each one.

fit(df: pandas.core.frame.DataFrame) → etna.transforms.decomposition.change_points_trend._OneSegmentChangePointsTrendTransform

Fit OneSegmentChangePointsTransform: find trend change points in df, fit detrend models with data from intervals of stable trend.

Parameters

df (pandas.core.frame.DataFrame) – one segment dataframe indexed with timestamp

Return type

etna.transforms.decomposition.change_points_trend._OneSegmentChangePointsTrendTransform

inverse_transform(df: pandas.core.frame.DataFrame) → pandas.core.frame.DataFrame

Split df to intervals of stable trend according to previous change point detection and add trend to each one.

Parameters

df (pandas.core.frame.DataFrame) – one segment dataframe to turn trend back

Returns

df – df with restored trend in in_column

Return type

pd.DataFrame

transform(df: pandas.core.frame.DataFrame) → pandas.core.frame.DataFrame

Split df to intervals of stable trend and subtract trend from each one.

Parameters

df (pandas.core.frame.DataFrame) – one segment dataframe to subtract trend

Returns

detrended df – df with detrended in_column series

Return type

pd.DataFrame