etna.transforms.FourierTransform#

class FourierTransform(period: float, order: int | None = None, mods: Sequence[int] | None = None, out_column: str | None = None, in_column: str | None = None)[source]#

Bases: IrreversibleTransform

Adds fourier features to the dataset.

Transform can work with two types of timestamp data: numeric and datetime.

Transform can accept timestamp data in two forms:

As index. In this case the dataset index is used to compute features. The features will be the same for each segment.
As external column. In this case for each segment its in_column will be used to compute features. It is expected that for each segment we have the same type of timestamp data (datetime or numeric), and for datetime type only one frequency is used for all the segments.

If we are working with external column, there is a difference in handling numeric and datetime data:

Numeric data can have missing values at any place.
Datetime data could have missing values only at the beginning of each segment.

Notes

To understand how transform works we recommend reading: Fourier series.

If we already have a numeric data then for a mode $m$ with a period $p$ we have:

\[\begin{split}& k = \left \lfloor \frac{m}{2} \right \rfloor \\ & f_{m, i} = \sin \left( \frac{2 \pi k i}{p} + \frac{\pi}{2} (m \mod 2) \right)\end{split}\]

If we have datetime data, then it first should be transformed into numeric. During fitting the transform saves frequency and some datetime timestamp as a reference point. During transformation it uses reference point to compute number of frequency units between reference point and each timestamp.

Create instance of FourierTransform.

Parameters:

period (float) –
the period of the seasonality to capture in frequency units of time series;

period should be >= 2
order (int | None) –
upper order of Fourier components to include;

order should be >= 1 and <= ceil(period/2))
mods (Sequence[int] | None) –
alternative and precise way of defining which harmonics will be used, for example, order=2 can be represented as mods=[1, 2, 3, 4] if period > 4 and as mods=[1, 2, 3] if 3 <= period <= 4.

mods should be >= 1 and < period
out_column (str | None) –
- if set, name of added column, the final name will be ‘{out_columnt}_{mod}’;
- if don’t set, name will be transform.__repr__(), repr will be made for transform that creates exactly this column
in_column (str | None) –
name of column to work with:
- if in_column is None (default) both datetime and integer timestamps are supported;
- if in_column isn’t None datetime and numeric columns are supported, but for datetime values only regular timestamps with some frequency are supported

Raises:

ValueError: – if period < 2
ValueError: – if both or none of order, mods is set
ValueError: – if order is < 1 or > ceil(period/2)
ValueError: – if at least one mod is < 1 or >= period

Methods

`fit`(ts)	Fit the transform.
`fit_transform`(ts)	Fit and transform TSDataset.
`get_regressors_info`()	Return the list with regressors created by the transform.
`inverse_transform`(ts)	Inverse transform TSDataset.
`load`(path)	Load an object.
`params_to_tune`()	Get default grid for tuning hyperparameters.
`save`(path)	Save the object.
`set_params`(**params)	Return new object instance with modified parameters.
`to_dict`()	Collect all information about etna object in dict.
`transform`(ts)	Transform TSDataset inplace.

Attributes

This class stores its __init__ parameters as attributes.

fit(ts: TSDataset) → FourierTransform[source]#

Fit the transform.

Parameters:

ts (TSDataset) – Dataset to fit the transform on.

Returns:

The fitted transform instance.

Raises:

ValueError – if external timestamp doesn’t have frequency
ValueError – if external timestamp doesn’t have the same frequency for all segments

Return type:

FourierTransform

fit_transform(ts: TSDataset) → TSDataset[source]#

Fit and transform TSDataset.

May be reimplemented. But it is not recommended.

Parameters:: ts (TSDataset) – TSDataset to transform.
Returns:: Transformed TSDataset.
Return type:: TSDataset

get_regressors_info() → List[str][source]#

Return the list with regressors created by the transform.

Return type:: List[str]

inverse_transform(ts: TSDataset) → TSDataset[source]#

Inverse transform TSDataset.

Do nothing.

Parameters:: ts (TSDataset) – TSDataset to be inverse transformed.
Returns:: TSDataset after applying inverse transformation.
Return type:: TSDataset

classmethod load(path: Path) → Self[source]#

Load an object.

Warning

This method uses dill module which is not secure. It is possible to construct malicious data which will execute arbitrary code during loading. Never load data that could have come from an untrusted source, or that could have been tampered with.

Parameters:: path (Path) – Path to load object from.
Returns:: Loaded object.
Return type:: Self

params_to_tune() → Dict[str, BaseDistribution][source]#

Get default grid for tuning hyperparameters.

If self.order is set then this grid tunes order parameter: Other parameters are expected to be set by the user.

Returns:: Grid to tune.
Return type:: Dict[str, BaseDistribution]

save(path: Path)[source]#

Save the object.

Parameters:: path (Path) – Path to save object to.

set_params(**params: dict) → Self[source]#

Return new object instance with modified parameters.

Method also allows to change parameters of nested objects within the current object. For example, it is possible to change parameters of a model in a Pipeline.

Nested parameters are expected to be in a <component_1>.<...>.<parameter> form, where components are separated by a dot.

Parameters:: **params (dict) – Estimator parameters
Returns:: New instance with changed parameters
Return type:: Self

Examples

>>> from etna.pipeline import Pipeline
>>> from etna.models import NaiveModel
>>> from etna.transforms import AddConstTransform
>>> model = NaiveModel(lag=1)
>>> transforms = [AddConstTransform(in_column="target", value=1)]
>>> pipeline = Pipeline(model, transforms=transforms, horizon=3)
>>> pipeline.set_params(**{"model.lag": 3, "transforms.0.value": 2})
Pipeline(model = NaiveModel(lag = 3, ), transforms = [AddConstTransform(in_column = 'target', value = 2, inplace = True, out_column = None, )], horizon = 3, )

to_dict()[source]#: Collect all information about etna object in dict.

transform(ts: TSDataset) → TSDataset[source]#

Transform TSDataset inplace.

Parameters:: ts (TSDataset) – Dataset to transform.
Returns:: Transformed TSDataset.
Return type:: TSDataset