Skip to content

Preprocessing

Classes:

Name Description
MinMaxScaler

Scale data using min-max scaling.
OneHotEncoder

One-hot encodes data.
StandardScaler

summary

MinMaxScaler

Scale data using min-max scaling.

Parameters:

Name Type Description Default
feature_range tuple[float, float]

The range to scale the data to, by default (0, 1)

 (0, 1)
Added in version 0.1.0

Methods:

Name Description
fit

summary
load

summary
save

summary
Source code in python/rapidstats/preprocessing.py 
 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
201
202
203
class MinMaxScaler:
    """Scale data using min-max scaling.

    Parameters
    ----------
    feature_range : tuple[float, float], optional
        The range to scale the data to, by default (0, 1)

    Added in version 0.1.0
    ----------------------
    """

    def __init__(self, feature_range: tuple[float, float] = (0, 1)):
        self.feature_range = feature_range
        self._set_range_vars()

    def _set_range_vars(self):
        self._range_min, self._range_max = self.feature_range
        self._range_diff = self._range_max - self._range_min

        return self

    def fit(self, X: nwt.IntoDataFrameT, columns: Optional[str | Iterable[str]] = None):
        """_summary_

        Parameters
        ----------
        X : nwt.IntoDataFrameT
            _description_

        Attributes
        ----------
        feature_names_in : list[str]
        min_: nwt.DataFrameT
        scale_ : nwt.DataFrameT

        Returns
        -------
        self
            Fitted MinMaxScaler
        """
        X = nw.from_native(X, eager_only=True)

        self.feature_names_in_ = _resolve_columns(X, columns)
        data_min = X.select(nw.col(self.feature_names_in_).min())
        data_max = X.select(nw.col(self.feature_names_in_).max())
        data_range: nwt.DataFrameT = data_max.select(
            nw.col(c).__sub__(data_min[c]) for c in self.feature_names_in_
        )

        self.scale_ = data_range.with_columns(
            nw.lit(self._range_diff).__truediv__(nw.col(c)).alias(c)
            for c in self.feature_names_in_
        )

        self.min_ = data_min.select(
            nw.lit(self._range_min).__sub__(nw.col(c).__mul__(self.scale_[c])).alias(c)
            for c in self.feature_names_in_
        )

        return self

    @nw.narwhalify
    def transform(self, X: nwt.IntoFrameT) -> nwt.IntoFrameT:
        return X.with_columns(
            nw.col(c).__mul__(self.scale_[c]).__add__(self.min_[c])
            for c in self.feature_names_in_
        )

    @nw.narwhalify
    def fit_transform(self, X: nwt.IntoDataFrameT) -> nwt.IntoDataFrameT:
        return self.fit(X).transform(X)

    @nw.narwhalify
    def inverse_transform(self, X: nwt.IntoFrameT) -> nwt.IntoFrameT:
        return X.with_columns(
            nw.col(c).__sub__(self.min_[c]).__truediv__(self.scale_[c])
            for c in self.feature_names_in_
        )

    def _run_one(self, c: str) -> nw.Expr:
        expr = nw.col(c)
        min_ = expr.min()

        return (
            expr.__sub__(min_)
            .__truediv__(expr.max().__sub__(min_))
            .__mul__(self._range_diff)
            .__add__(self._range_min)
        )

    @nw.narwhalify
    def run(
        self, X: nwt.IntoFrameT, columns: Optional[str | Iterable[str]] = None
    ) -> nwt.IntoFrameT:
        return X.with_columns(self._run_one(c) for c in _resolve_columns(X, columns))

    def save(self, path: PathLike):
        """_summary_

        Parameters
        ----------
        path : PathLike
            _description_

        Returns
        -------
        _type_
            _description_

        Added in version 0.2.0
        ----------------------
        """
        with zipfile.ZipFile(
            path, "w"
        ) as archive, tempfile.TemporaryDirectory() as tmpdir:
            tmpdir = Path(tmpdir)

            self.min_.write_parquet(tmpdir / "min_.parquet")
            self.scale_.write_parquet(tmpdir / "scale_.parquet")
            _write_json(
                {
                    "feature_names_in_": self.feature_names_in_,
                    "feature_range": self.feature_range,
                },
                tmpdir / "instance_vars.json",
            )

            archive.write(tmpdir / "min_.parquet", "min_.parquet")
            archive.write(tmpdir / "scale_.parquet", "scale_.parquet")
            archive.write(tmpdir / "instance_vars.json", "instance_vars.json")

        return self

    def load(self, path: PathLike):
        """_summary_

        Parameters
        ----------
        path : PathLike
            _description_

        Returns
        -------
        _type_
            _description_

        Added in version 0.2.0
        ----------------------
        """
        with zipfile.ZipFile(
            path, "r"
        ) as archive, tempfile.TemporaryDirectory() as tmpdir:
            archive.extractall(tmpdir)

            self.min_ = nw.read_parquet(f"{tmpdir}/min_.parquet", native_namespace=pl)
            self.scale_ = nw.read_parquet(
                f"{tmpdir}/scale_.parquet", native_namespace=pl
            )
            instance_vars = _read_json(f"{tmpdir}/instance_vars.json")
            self.feature_names_in_ = instance_vars["feature_names_in_"]
            self.feature_range = tuple(instance_vars["feature_range"])

            self._set_range_vars()

        return self

fit(X, columns=None)

summary

Parameters:

Name Type Description Default
X IntoDataFrameT

description

 required

Attributes:

Name Type Description
feature_names_in list[str]
min_ DataFrameT
scale_ DataFrameT

Returns:

Type Description
self

Fitted MinMaxScaler
Source code in python/rapidstats/preprocessing.py 
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
def fit(self, X: nwt.IntoDataFrameT, columns: Optional[str | Iterable[str]] = None):
    """_summary_

    Parameters
    ----------
    X : nwt.IntoDataFrameT
        _description_

    Attributes
    ----------
    feature_names_in : list[str]
    min_: nwt.DataFrameT
    scale_ : nwt.DataFrameT

    Returns
    -------
    self
        Fitted MinMaxScaler
    """
    X = nw.from_native(X, eager_only=True)

    self.feature_names_in_ = _resolve_columns(X, columns)
    data_min = X.select(nw.col(self.feature_names_in_).min())
    data_max = X.select(nw.col(self.feature_names_in_).max())
    data_range: nwt.DataFrameT = data_max.select(
        nw.col(c).__sub__(data_min[c]) for c in self.feature_names_in_
    )

    self.scale_ = data_range.with_columns(
        nw.lit(self._range_diff).__truediv__(nw.col(c)).alias(c)
        for c in self.feature_names_in_
    )

    self.min_ = data_min.select(
        nw.lit(self._range_min).__sub__(nw.col(c).__mul__(self.scale_[c])).alias(c)
        for c in self.feature_names_in_
    )

    return self

load(path)

summary

Parameters:

Name Type Description Default
path PathLike

description

 required

Returns:

Type Description
_type_

description
Added in version 0.2.0
Source code in python/rapidstats/preprocessing.py 
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
201
202
203
def load(self, path: PathLike):
    """_summary_

    Parameters
    ----------
    path : PathLike
        _description_

    Returns
    -------
    _type_
        _description_

    Added in version 0.2.0
    ----------------------
    """
    with zipfile.ZipFile(
        path, "r"
    ) as archive, tempfile.TemporaryDirectory() as tmpdir:
        archive.extractall(tmpdir)

        self.min_ = nw.read_parquet(f"{tmpdir}/min_.parquet", native_namespace=pl)
        self.scale_ = nw.read_parquet(
            f"{tmpdir}/scale_.parquet", native_namespace=pl
        )
        instance_vars = _read_json(f"{tmpdir}/instance_vars.json")
        self.feature_names_in_ = instance_vars["feature_names_in_"]
        self.feature_range = tuple(instance_vars["feature_range"])

        self._set_range_vars()

    return self

save(path)

summary

Parameters:

Name Type Description Default
path PathLike

description

 required

Returns:

Type Description
_type_

description
Added in version 0.2.0
Source code in python/rapidstats/preprocessing.py 
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
def save(self, path: PathLike):
    """_summary_

    Parameters
    ----------
    path : PathLike
        _description_

    Returns
    -------
    _type_
        _description_

    Added in version 0.2.0
    ----------------------
    """
    with zipfile.ZipFile(
        path, "w"
    ) as archive, tempfile.TemporaryDirectory() as tmpdir:
        tmpdir = Path(tmpdir)

        self.min_.write_parquet(tmpdir / "min_.parquet")
        self.scale_.write_parquet(tmpdir / "scale_.parquet")
        _write_json(
            {
                "feature_names_in_": self.feature_names_in_,
                "feature_range": self.feature_range,
            },
            tmpdir / "instance_vars.json",
        )

        archive.write(tmpdir / "min_.parquet", "min_.parquet")
        archive.write(tmpdir / "scale_.parquet", "scale_.parquet")
        archive.write(tmpdir / "instance_vars.json", "instance_vars.json")

    return self

OneHotEncoder

One-hot encodes data.

Added in version 0.1.0

Methods:

Name Description
load

summary
Source code in python/rapidstats/preprocessing.py 
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
class OneHotEncoder:
    """One-hot encodes data.

    Added in version 0.1.0
    ----------------------
    """

    def __init__(self):
        pass

    def fit(self, X: nwt.IntoDataFrameT, columns: Optional[str | Iterable[str]] = None):
        X = nw.from_native(X, eager_only=True)

        self.categories_ = {
            c: X[c].drop_nulls().unique() for c in _resolve_columns(X, columns)
        }

        return self

    @nw.narwhalify
    def transform(self, X: nwt.IntoFrameT) -> nwt.IntoFrameT:
        for c, unique_vals in self.categories_.items():
            for val in unique_vals:
                X = X.with_columns(nw.col(c).__eq__(val).alias(f"{c}_{val}"))

        return X

    @nw.narwhalify
    def fit_transform(
        self, X: nwt.IntoDataFrameT, columns: Optional[str | Iterable[str]] = None
    ) -> nwt.IntoDataFrameT:
        return self.fit(X, columns=columns).transform(X)

    def save(self, path: PathLike):
        with zipfile.ZipFile(
            path, "w"
        ) as archive, tempfile.TemporaryDirectory() as tmpdir:
            tmpdir = Path(tmpdir)

            for k, v in self.categories_.items():
                v.to_frame().write_parquet(tmpdir / k)
                archive.write(tmpdir / k, k)

        return self

    def load(self, path: PathLike):
        """_summary_

        Parameters
        ----------
        path : PathLike
            _description_

        Returns
        -------
        Self
            _description_

        Added in version 0.2.0
        ----------------------
        """
        with zipfile.ZipFile(
            path, "r"
        ) as archive, tempfile.TemporaryDirectory() as tmpdir:
            archive.extractall(tmpdir)

            self.categories_ = {
                file: nw.read_parquet(f"{tmpdir}/{file}", native_namespace=pl)[file]
                for file in archive.namelist()
            }

        return self

load(path)

summary

Parameters:

Name Type Description Default
path PathLike

description

 required

Returns:

Type Description
Self

description
Added in version 0.2.0
Source code in python/rapidstats/preprocessing.py 
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
def load(self, path: PathLike):
    """_summary_

    Parameters
    ----------
    path : PathLike
        _description_

    Returns
    -------
    Self
        _description_

    Added in version 0.2.0
    ----------------------
    """
    with zipfile.ZipFile(
        path, "r"
    ) as archive, tempfile.TemporaryDirectory() as tmpdir:
        archive.extractall(tmpdir)

        self.categories_ = {
            file: nw.read_parquet(f"{tmpdir}/{file}", native_namespace=pl)[file]
            for file in archive.namelist()
        }

    return self

StandardScaler

summary

Null

rapidstats uses narwhals to ingest supported DataFrames. However, null-handling can differ across backends. For example, if using a Polars backend, NaNs are valid numbers, not missing. Therefore, the mean / standard deviation of a column with NaNs will be NaN. Ensure your input is sanitized according to your specific backend before using StandardScaler.

Source code in python/rapidstats/preprocessing.py 
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
class StandardScaler:
    """_summary_

    !!! Null Handling
        `rapidstats` uses [narwhals](https://narwhals-dev.github.io/narwhals/) to ingest
        supported DataFrames. However, null-handling can differ across backends. For
        example, if using a Polars backend, NaNs are valid numbers, not missing.
        Therefore, the mean / standard deviation of a column with NaNs will be NaN.
        Ensure your input is sanitized according to your specific backend before using
        `StandardScaler`.
    """

    def __init__(self, ddof: int = 1):
        self.ddof = ddof

    def fit(self, X: nwt.IntoDataFrame, columns: Optional[str | Iterable[str]] = None):
        X = nw.from_native(X, eager_only=True)
        self.feature_names_in_ = _resolve_columns(X, columns)
        selector = nw.col(self.feature_names_in_)

        self.mean_ = X.select(selector.mean())
        self.std_ = X.select(selector.std(ddof=self.ddof))

        return self

    @nw.narwhalify
    def transform(self, X: nwt.IntoFrameT) -> nwt.IntoFrameT:
        return X.with_columns(
            nw.col(c).__sub__(self.mean_[c]).__truediv__(self.std_[c])
            for c in self.feature_names_in_
        )

    @nw.narwhalify
    def fit_transform(
        self, X: nwt.IntoDataFrameT, columns: Optional[str | Iterable[str]] = None
    ) -> nwt.IntoDataFrameT:
        return self.fit(X, columns=columns).transform(X)

    @nw.narwhalify
    def inverse_transform(self, X: nwt.IntoFrameT) -> nwt.IntoFrameT:
        return X.with_columns(
            nw.col(c).__mul__(self.std_[c]).__add__(self.mean_[c])
            for c in self.feature_names_in_
        )

    def _run_one(self, c: str) -> nw.Expr:
        expr = nw.col(c)

        return expr.__sub__(expr.mean()).__truediv__(expr.std(ddof=self.ddof))

    @nw.narwhalify
    def run(
        self, X: nwt.IntoFrameT, columns: Optional[str | Iterable[str]] = None
    ) -> nwt.IntoFrameT:
        return X.with_columns(self._run_one(c) for c in _resolve_columns(X, columns))

    def save(self, path: PathLike):
        with zipfile.ZipFile(
            path, "w"
        ) as archive, tempfile.TemporaryDirectory() as tmpdir:
            tmpdir = Path(tmpdir)

            self.mean_.write_parquet(tmpdir / "mean_.parquet")
            self.std_.write_parquet(tmpdir / "std_.parquet")
            _write_json(
                {
                    "feature_names_in_": self.feature_names_in_,
                    "ddof": self.ddof,
                },
                tmpdir / "instance_vars.json",
            )

            archive.write(tmpdir / "mean_.parquet", "mean_.parquet")
            archive.write(tmpdir / "std_.parquet", "std_.parquet")
            archive.write(tmpdir / "instance_vars.json", "instance_vars.json")

        return self

    def load(self, path: PathLike):
        with zipfile.ZipFile(
            path, "r"
        ) as archive, tempfile.TemporaryDirectory() as tmpdir:
            archive.extractall(tmpdir)

            self.mean_ = nw.read_parquet(f"{tmpdir}/mean_.parquet", native_namespace=pl)
            self.std_ = nw.read_parquet(f"{tmpdir}/std_.parquet", native_namespace=pl)
            instance_vars = _read_json(f"{tmpdir}/instance_vars.json")
            self.feature_names_in_ = instance_vars["feature_names_in_"]
            self.ddof = instance_vars["ddof"]

        return self