Back to models
Transformer

ScaledAsinhTransformer

API referenceSource
MultivariateInverse transformUnequal-length series

Hyperbolic sine transformation and its inverse [1].

Known as variance stabilizing transformation, Combined with an sktime.forecasting.compose.TransformedTargetForecaster, can be useful in time series that exhibit spikes [1], [2]

Quickstart

python
from sktime.transformations.scaledasinh import ScaledAsinhTransformer

estimator = ScaledAsinhTransformer(mad_normalization_factor=1.4826)

Tags

Capabilities

  • Multivariate
  • Inverse transform: Supported
  • Unequal-length series
  • Missing values: Not supported
  • Removes missing values: Not supported
  • Equalizes series length: Not supported

Properties

Input typescitype:transform-input
Series
Output typescitype:transform-output
Series
Label typescitype:transform-labels
None
Fit is emptyfit_is_empty
No
Keeps the time indextransform-returns-same-time-index
Yes
Requires Xrequires_X
Yes
Requires yrequires_y
No
X and y need the same indexX-y-must-have-same-index
No

Parameters(1)

mad_normalization_factorfloat, default = 1.4826

The normalization factor used to adjust the median absolute deviation (MAD) for asymptotically normal consistency to the standard deviation. The default value based on [1], [2] is 1.4826.

Examples

>>> from sktime.transformations.scaledasinh import ScaledAsinhTransformer
>>> from sktime.datasets import load_airline
>>> y = load_airline ()
>>> transformer = ScaledAsinhTransformer ()
>>> y_hat = transformer. fit_transform (y)

References

[1] (1,2,3,4,5,6,7)

Ziel F, Weron R. Day-ahead electricity price forecasting with high-dimensional structures: Univariate vs. multivariate modeling frameworks. Energy Economics. 2018 Feb 1;70:396-420.

[2] (1,2,3)

Uniejewski, B., Weron, R., Ziel, F., 2017. Variance stabilizing transformations for electricity spot price forecasting. IEEE Transactions on Power Systems, DOI: 10.1109/TPWRS.2017.2734563