MaxAbsScaler¶

class MaxAbsScaler(SklearnScaler):
    """Scale each feature by its maximum absolute value.

    This estimator scales and translates each feature individually such that
    the maximal absolute value of each feature in the training set will be 1.0.
    It does not shift/center the data, and thus does not destroy any sparsity.

    ``MaxAbsScaler`` doesn't reduce the effect of outliers; it only linearly
    scales them down.

    This is a Yohou wrapper that preserves the polars DataFrame structure and
    "time" column.

    Parameters
    ----------
    copy : bool, default=True
        If True, a copy of the data is made before transforming. Pass False only
        if in-place modification of the input is acceptable.

    clip : bool, default=False
        Set to True to clip transformed values of held-out data to [-1, 1].

    Attributes
    ----------
    instance_ : sklearn.preprocessing.MaxAbsScaler
        The fitted sklearn MaxAbsScaler instance.

    scale_ : ndarray of shape (n_features,)
        Per feature relative scaling of the data.

    max_abs_ : ndarray of shape (n_features,)
        Per feature maximum absolute value.

    Examples
    --------
    >>> import polars as pl
    >>> from datetime import datetime
    >>> from yohou.preprocessing import MaxAbsScaler
    >>> X = pl.DataFrame({
    ...     "time": [datetime(2024, 1, i) for i in range(1, 6)],
    ...     "value": [-50.0, -25.0, 0.0, 25.0, 50.0],
    ... })
    >>> scaler = MaxAbsScaler()
    >>> scaler.fit(X)  # doctest: +ELLIPSIS
    MaxAbsScaler(...)
    >>> X_scaled = scaler.transform(X)
    >>> # Values scaled to [-1, 1] range
    >>> X_scaled["value"].max()
    1.0
    >>> X_scaled["value"].min()
    -1.0

    See Also
    --------
    - [`MinMaxScaler`][yohou.preprocessing.sklearn_wrappers.MinMaxScaler] : Scale features to a given range.
    - [`StandardScaler`][yohou.preprocessing.sklearn_wrappers.StandardScaler] : Standardize features by removing mean and scaling to unit variance.

    """

    _estimator_default_class = sklearn_MaxAbsScaler

    def __init__(self, copy=True, clip=False, **kwargs):
        # ``clip`` was added to MaxAbsScaler in sklearn 1.7; drop it when running
        # against an older sklearn that does not accept it, which otherwise raises.
        params = _filter_estimator_params(sklearn_MaxAbsScaler, {"copy": copy, "clip": clip})
        super().__init__(**params, **kwargs)

    @property
    def scale_(self) -> np.ndarray:
        """Per feature relative scaling of the data."""
        check_is_fitted(self, ["instance_"])
        return self.instance_.scale_

    @property
    def max_abs_(self) -> np.ndarray:
        """Per feature maximum absolute value."""
        check_is_fitted(self, ["instance_"])
        return self.instance_.max_abs_