Module model_selection (1.30.0)

Functions for test/train split and model tuning. This module is styled after scikit-learn's model_selection module: https://scikit-learn.org/stable/modules/classes.html#module-sklearn.model_selection.

Classes

KFold

KFold(n_splits: int = 5, *, random_state: typing.Optional[int] = None)

K-Fold cross-validator.

Split data in train/test sets. Split dataset into k consecutive folds.

Each fold is then used once as a validation while the k - 1 remaining folds form the training set.

Examples:

>>> import bigframes.pandas as bpd
>>> from bigframes.ml.model_selection import KFold
>>> bpd.options.display.progress_bar = None
>>> X = bpd.DataFrame({"feat0": [1, 3, 5], "feat1": [2, 4, 6]})
>>> y = bpd.DataFrame({"label": [1, 2, 3]})
>>> kf = KFold(n_splits=3, random_state=42)
>>> for i, (X_train, X_test, y_train, y_test) in enumerate(kf.split(X, y)):
...     print(f"Fold {i}:")
...     print(f"  X_train: {X_train}")
...     print(f"  X_test: {X_test}")
...     print(f"  y_train: {y_train}")
...     print(f"  y_test: {y_test}")
...
Fold 0:
  X_train:    feat0  feat1
1      3      4
2      5      6
<BLANKLINE>
[2 rows x 2 columns]
  X_test:    feat0  feat1
0      1      2
<BLANKLINE>
[1 rows x 2 columns]
  y_train:    label
1      2
2      3
<BLANKLINE>
[2 rows x 1 columns]
  y_test:    label
0      1
<BLANKLINE>
[1 rows x 1 columns]
Fold 1:
  X_train:    feat0  feat1
0      1      2
2      5      6
<BLANKLINE>
[2 rows x 2 columns]
  X_test:    feat0  feat1
1      3      4
<BLANKLINE>
[1 rows x 2 columns]
  y_train:    label
0      1
2      3
<BLANKLINE>
[2 rows x 1 columns]
  y_test:    label
1      2
<BLANKLINE>
[1 rows x 1 columns]
Fold 2:
  X_train:    feat0  feat1
0      1      2
1      3      4
<BLANKLINE>
[2 rows x 2 columns]
  X_test:    feat0  feat1
2      5      6
<BLANKLINE>
[1 rows x 2 columns]
  y_train:    label
0      1
1      2
<BLANKLINE>
[2 rows x 1 columns]
  y_test:    label
2      3
<BLANKLINE>
[1 rows x 1 columns]

Parameters
Name	Description
`n_splits`	`int` Number of folds. Must be at least 2. Default to 5.
`random_state`	`Optional[int]` A seed to use for randomly choosing the rows of the split. If not set, a random split will be generated each time. Default to None.

Modules Functions

cross_validate

cross_validate(
    estimator,
    X: typing.Union[
        bigframes.dataframe.DataFrame,
        bigframes.series.Series,
        pandas.core.frame.DataFrame,
        pandas.core.series.Series,
    ],
    y: typing.Optional[
        typing.Union[
            bigframes.dataframe.DataFrame,
            bigframes.series.Series,
            pandas.core.frame.DataFrame,
            pandas.core.series.Series,
        ]
    ] = None,
    *,
    cv: typing.Optional[typing.Union[int, bigframes.ml.model_selection.KFold]] = None
) -> dict[str, list]

Evaluate metric(s) by cross-validation and also record fit/score times.

Examples:

>>> import bigframes.pandas as bpd
>>> from bigframes.ml.model_selection import cross_validate, KFold
>>> from bigframes.ml.linear_model import LinearRegression
>>> bpd.options.display.progress_bar = None
>>> X = bpd.DataFrame({"feat0": [1, 3, 5], "feat1": [2, 4, 6]})
>>> y = bpd.DataFrame({"label": [1, 2, 3]})
>>> model = LinearRegression()
>>> scores = cross_validate(model, X, y, cv=3) # doctest: +SKIP
>>> for score in scores["test_score"]: # doctest: +SKIP
...   print(score["mean_squared_error"][0])
...
5.218167286047954e-19
2.726229944928669e-18
1.6197635612324266e-17

Parameters
Name	Description
`X`	`bigframes.dataframe.DataFrame or bigframes.series.Series` The data to fit.
`y`	`bigframes.dataframe.DataFrame, bigframes.series.Series or None` The target variable to try to predict in the case of supe()rvised learning. Default to None.
`cv`	`int, bigframes.ml.model_selection.KFold or None` Determines the cross-validation splitting strategy. Possible inputs for cv are: - None, to use the default 5-fold cross validation, - int, to specify the number of folds in a `KFold`, - bigframes.ml.model_selection.KFold instance.

Returns
Type	Description
`Dict[str, List]`	A dict of arrays containing the score/time arrays for each scorer is returned. The keys for this `dict` are: `test_score` The score array for test scores on each cv split. `fit_time` The time for fitting the estimator on the train set for each cv split. `score_time` The time for scoring the estimator on the test set for each cv split.

train_test_split

train_test_split(
    *arrays: typing.Union[
        bigframes.dataframe.DataFrame,
        bigframes.series.Series,
        pandas.core.frame.DataFrame,
        pandas.core.series.Series,
    ],
    test_size: typing.Optional[float] = None,
    train_size: typing.Optional[float] = None,
    random_state: typing.Optional[int] = None,
    stratify: typing.Optional[bigframes.series.Series] = None
) -> typing.List[typing.Union[bigframes.dataframe.DataFrame, bigframes.series.Series]]

Splits dataframes or series into random train and test subsets.

Examples:

>>> import bigframes.pandas as bpd
>>> from bigframes.ml.model_selection import train_test_split
>>> bpd.options.display.progress_bar = None
>>> X = bpd.DataFrame({"feat0": [0, 2, 4, 6, 8], "feat1": [1, 3, 5, 7, 9]})
>>> y = bpd.DataFrame({"label": [0, 1, 2, 3, 4]})
>>> X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=42)
>>> X_train
    feat0  feat1
0      0      1
1      2      3
4      8      9
<BLANKLINE>
[3 rows x 2 columns]
>>> y_train
    label
0      0
1      1
4      4
<BLANKLINE>
[3 rows x 1 columns]
>>> X_test
    feat0  feat1
2      4      5
3      6      7
<BLANKLINE>
[2 rows x 2 columns]
>>> y_test
    label
2      2
3      3
<BLANKLINE>
[2 rows x 1 columns]

Parameters
Name	Description
`\*arrays`	`bigframes.dataframe.DataFrame or bigframes.series.Series` A sequence of BigQuery DataFrames or Series that can be joined on their indexes.
`test_size`	`default None` The proportion of the dataset to include in the test split. If None, this will default to the complement of train_size. If both are none, it will be set to 0.25.
`train_size`	`default None` The proportion of the dataset to include in the train split. If None, this will default to the complement of test_size.
`random_state`	`default None` A seed to use for randomly choosing the rows of the split. If not set, a random split will be generated each time.

Returns
Type	Description
`List[Union[bigframes.dataframe.DataFrame, bigframes.series.Series]]`	A list of BigQuery DataFrames or Series.