Skip to content

RobustScaler

Bases: InvertibleTableTransformer

The RobustScaler transforms column values to a range by removing the median and scaling to the interquartile range.

Currently, for columns with high stability (IQR == 0), it will only subtract the median and not scale to avoid dividing by zero.

Parameters:

Name Type Description Default
selector str | list[str] | None

The list of columns used to fit the transformer. If None, all numeric columns are used.

 None

Methods:

Name Description
fit

Learn a transformation for a set of columns in a table.
fit_and_transform

Learn a transformation for a set of columns in a table and apply the learned transformation to the same table.
inverse_transform

Undo the learned transformation.
transform

Apply the learned transformation to a table.

Attributes:

Name Type Description
is_fitted bool

Whether the transformer is fitted.
Source code in src/safeds/data/tabular/transformation/_robust_scaler.py 
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
class RobustScaler(InvertibleTableTransformer):
    """
    The RobustScaler transforms column values to a range by removing the median and scaling to the interquartile range.

    Currently, for columns with high stability (IQR == 0), it will only subtract the median and not scale to avoid
    dividing by zero.

    Parameters
    ----------
    selector:
        The list of columns used to fit the transformer. If `None`, all numeric columns are used.
    """

    # ------------------------------------------------------------------------------------------------------------------
    # Dunder methods
    # ------------------------------------------------------------------------------------------------------------------

    def __init__(self, *, selector: str | list[str] | None = None) -> None:
        super().__init__(selector)

        # Internal state
        self._data_median: pl.DataFrame | None = None
        self._data_scale: pl.DataFrame | None = None

    def __hash__(self) -> int:
        # Leave out the internal state for faster hashing
        return super().__hash__()

    # ------------------------------------------------------------------------------------------------------------------
    # Properties
    # ------------------------------------------------------------------------------------------------------------------

    @property
    def is_fitted(self) -> bool:
        """Whether the transformer is fitted."""
        return self._data_median is not None and self._data_scale is not None

    # ------------------------------------------------------------------------------------------------------------------
    # Learning and transformation
    # ------------------------------------------------------------------------------------------------------------------

    def fit(self, table: Table) -> RobustScaler:
        """
        Learn a transformation for a set of columns in a table.

        **Note:** This transformer is not modified.

        Parameters
        ----------
        table:
            The table used to fit the transformer.

        Returns
        -------
        fitted_transformer:
            The fitted transformer.

        Raises
        ------
        ColumnNotFoundError
            If column_names contain a column name that is missing in the table.
        ColumnTypeError
            If at least one of the specified columns in the table contains non-numerical data.
        ValueError
            If the table contains 0 rows.
        """
        import polars as pl

        if self._selector is None:
            column_names = [name for name in table.column_names if table.get_column_type(name).is_numeric]
        else:
            column_names = self._selector
            _check_columns_exist(table, column_names)
            _check_columns_are_numeric(table, column_names, operation="fit a RobustScaler")

        if table.row_count == 0:
            raise ValueError("The RobustScaler cannot be fitted because the table contains 0 rows")

        _data_median = _safe_collect_lazy_frame(table._lazy_frame.select(column_names).median())
        q1 = _safe_collect_lazy_frame(table._lazy_frame.select(column_names).quantile(0.25))
        q3 = _safe_collect_lazy_frame(table._lazy_frame.select(column_names).quantile(0.75))
        _data_scale = q3 - q1

        # To make sure there is no division by zero
        for col_e in column_names:
            _data_scale = _data_scale.with_columns(
                pl.when(pl.col(col_e) == 0).then(1).otherwise(pl.col(col_e)).alias(col_e),
            )

        # Create a copy with the learned transformation
        result = RobustScaler(selector=column_names)
        result._data_median = _data_median
        result._data_scale = _data_scale

        return result

    def transform(self, table: Table) -> Table:
        """
        Apply the learned transformation to a table.

        **Note:** The given table is not modified.

        Parameters
        ----------
        table:
            The table to which the learned transformation is applied.

        Returns
        -------
        transformed_table:
            The transformed table.

        Raises
        ------
        NotFittedError
            If the transformer has not been fitted yet.
        ColumnNotFoundError
            If the input table does not contain all columns used to fit the transformer.
        ColumnTypeError
            If at least one of the columns in the input table that is used to fit contains non-numerical data.
        """
        import polars as pl

        # Used in favor of is_fitted, so the type checker is happy
        if self._selector is None or self._data_median is None or self._data_scale is None:
            raise NotFittedError(kind="transformer")

        _check_columns_exist(table, self._selector)
        _check_columns_are_numeric(table, self._selector, operation="transform with a RobustScaler")

        columns = [
            (pl.col(name) - self._data_median.get_column(name)) / self._data_scale.get_column(name)
            for name in self._selector
        ]

        return Table._from_polars_lazy_frame(
            table._lazy_frame.with_columns(columns),
        )

    def inverse_transform(self, transformed_table: Table) -> Table:
        """
        Undo the learned transformation.

        **Note:** The given table is not modified.

        Parameters
        ----------
        transformed_table:
            The table to be transformed back to the original version.

        Returns
        -------
        original_table:
            The original table.

        Raises
        ------
        NotFittedError
            If the transformer has not been fitted yet.
        ColumnNotFoundError
            If the input table does not contain all columns used to fit the transformer.
        ColumnTypeError
            If the transformed columns of the input table contain non-numerical data.
        """
        import polars as pl

        # Used in favor of is_fitted, so the type checker is happy
        if self._selector is None or self._data_median is None or self._data_scale is None:
            raise NotFittedError(kind="transformer")

        _check_columns_exist(transformed_table, self._selector)
        _check_columns_are_numeric(
            transformed_table,
            self._selector,
            operation="inverse-transform with a RobustScaler",
        )

        columns = [
            pl.col(name) * self._data_scale.get_column(name) + self._data_median.get_column(name)
            for name in self._selector
        ]

        return Table._from_polars_lazy_frame(
            transformed_table._lazy_frame.with_columns(columns),
        )

is_fitted

Whether the transformer is fitted.

fit

Learn a transformation for a set of columns in a table.

Note: This transformer is not modified.

Parameters:

Name Type Description Default
table Table

The table used to fit the transformer.

 required

Returns:

Name Type Description
fitted_transformer RobustScaler

The fitted transformer.

Raises:

Type Description
ColumnNotFoundError

If column_names contain a column name that is missing in the table.
ColumnTypeError

If at least one of the specified columns in the table contains non-numerical data.
ValueError

If the table contains 0 rows.
Source code in src/safeds/data/tabular/transformation/_robust_scaler.py 
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
def fit(self, table: Table) -> RobustScaler:
    """
    Learn a transformation for a set of columns in a table.

    **Note:** This transformer is not modified.

    Parameters
    ----------
    table:
        The table used to fit the transformer.

    Returns
    -------
    fitted_transformer:
        The fitted transformer.

    Raises
    ------
    ColumnNotFoundError
        If column_names contain a column name that is missing in the table.
    ColumnTypeError
        If at least one of the specified columns in the table contains non-numerical data.
    ValueError
        If the table contains 0 rows.
    """
    import polars as pl

    if self._selector is None:
        column_names = [name for name in table.column_names if table.get_column_type(name).is_numeric]
    else:
        column_names = self._selector
        _check_columns_exist(table, column_names)
        _check_columns_are_numeric(table, column_names, operation="fit a RobustScaler")

    if table.row_count == 0:
        raise ValueError("The RobustScaler cannot be fitted because the table contains 0 rows")

    _data_median = _safe_collect_lazy_frame(table._lazy_frame.select(column_names).median())
    q1 = _safe_collect_lazy_frame(table._lazy_frame.select(column_names).quantile(0.25))
    q3 = _safe_collect_lazy_frame(table._lazy_frame.select(column_names).quantile(0.75))
    _data_scale = q3 - q1

    # To make sure there is no division by zero
    for col_e in column_names:
        _data_scale = _data_scale.with_columns(
            pl.when(pl.col(col_e) == 0).then(1).otherwise(pl.col(col_e)).alias(col_e),
        )

    # Create a copy with the learned transformation
    result = RobustScaler(selector=column_names)
    result._data_median = _data_median
    result._data_scale = _data_scale

    return result

fit_and_transform

Learn a transformation for a set of columns in a table and apply the learned transformation to the same table.

Note: Neither this transformer nor the given table are modified.

Parameters:

Name Type Description Default
table Table

The table used to fit the transformer. The transformer is then applied to this table.

 required

Returns:

Name Type Description
fitted_transformer Self

The fitted transformer.
transformed_table Table

The transformed table.
Source code in src/safeds/data/tabular/transformation/_table_transformer.py 
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
def fit_and_transform(self, table: Table) -> tuple[Self, Table]:
    """
    Learn a transformation for a set of columns in a table and apply the learned transformation to the same table.

    **Note:** Neither this transformer nor the given table are modified.

    Parameters
    ----------
    table:
        The table used to fit the transformer. The transformer is then applied to this table.

    Returns
    -------
    fitted_transformer:
        The fitted transformer.
    transformed_table:
        The transformed table.
    """
    fitted_transformer = self.fit(table)
    transformed_table = fitted_transformer.transform(table)
    return fitted_transformer, transformed_table

inverse_transform

Undo the learned transformation.

Note: The given table is not modified.

Parameters:

Name Type Description Default
transformed_table Table

The table to be transformed back to the original version.

 required

Returns:

Name Type Description
original_table Table

The original table.

Raises:

Type Description
NotFittedError

If the transformer has not been fitted yet.
ColumnNotFoundError

If the input table does not contain all columns used to fit the transformer.
ColumnTypeError

If the transformed columns of the input table contain non-numerical data.
Source code in src/safeds/data/tabular/transformation/_robust_scaler.py 
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
def inverse_transform(self, transformed_table: Table) -> Table:
    """
    Undo the learned transformation.

    **Note:** The given table is not modified.

    Parameters
    ----------
    transformed_table:
        The table to be transformed back to the original version.

    Returns
    -------
    original_table:
        The original table.

    Raises
    ------
    NotFittedError
        If the transformer has not been fitted yet.
    ColumnNotFoundError
        If the input table does not contain all columns used to fit the transformer.
    ColumnTypeError
        If the transformed columns of the input table contain non-numerical data.
    """
    import polars as pl

    # Used in favor of is_fitted, so the type checker is happy
    if self._selector is None or self._data_median is None or self._data_scale is None:
        raise NotFittedError(kind="transformer")

    _check_columns_exist(transformed_table, self._selector)
    _check_columns_are_numeric(
        transformed_table,
        self._selector,
        operation="inverse-transform with a RobustScaler",
    )

    columns = [
        pl.col(name) * self._data_scale.get_column(name) + self._data_median.get_column(name)
        for name in self._selector
    ]

    return Table._from_polars_lazy_frame(
        transformed_table._lazy_frame.with_columns(columns),
    )

transform

Apply the learned transformation to a table.

Note: The given table is not modified.

Parameters:

Name Type Description Default
table Table

The table to which the learned transformation is applied.

 required

Returns:

Name Type Description
transformed_table Table

The transformed table.

Raises:

Type Description
NotFittedError

If the transformer has not been fitted yet.
ColumnNotFoundError

If the input table does not contain all columns used to fit the transformer.
ColumnTypeError

If at least one of the columns in the input table that is used to fit contains non-numerical data.
Source code in src/safeds/data/tabular/transformation/_robust_scaler.py 
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
def transform(self, table: Table) -> Table:
    """
    Apply the learned transformation to a table.

    **Note:** The given table is not modified.

    Parameters
    ----------
    table:
        The table to which the learned transformation is applied.

    Returns
    -------
    transformed_table:
        The transformed table.

    Raises
    ------
    NotFittedError
        If the transformer has not been fitted yet.
    ColumnNotFoundError
        If the input table does not contain all columns used to fit the transformer.
    ColumnTypeError
        If at least one of the columns in the input table that is used to fit contains non-numerical data.
    """
    import polars as pl

    # Used in favor of is_fitted, so the type checker is happy
    if self._selector is None or self._data_median is None or self._data_scale is None:
        raise NotFittedError(kind="transformer")

    _check_columns_exist(table, self._selector)
    _check_columns_are_numeric(table, self._selector, operation="transform with a RobustScaler")

    columns = [
        (pl.col(name) - self._data_median.get_column(name)) / self._data_scale.get_column(name)
        for name in self._selector
    ]

    return Table._from_polars_lazy_frame(
        table._lazy_frame.with_columns(columns),
    )