Source code for sklearn.model_selection._split

"""
The :mod:`sklearn.model_selection._split` module includes classes and
functions to split the data based on a preset strategy.
"""

# Author: Alexandre Gramfort <[email protected]>
#         Gael Varoquaux <[email protected]>
#         Olivier Grisel <[email protected]>
#         Raghav RV <[email protected]>
#         Leandro Hermida <[email protected]>
#         Rodion Martynov <[email protected]>
# License: BSD 3 clause

from collections.abc import Iterable
from collections import defaultdict
import warnings
from itertools import chain, combinations
from math import ceil, floor
import numbers
from abc import ABCMeta, abstractmethod
from inspect import signature

import numpy as np
from scipy.special import comb

from ..utils import indexable, check_random_state, _safe_indexing
from ..utils import _approximate_mode
from ..utils.validation import _num_samples, column_or_1d
from ..utils.validation import check_array
from ..utils.multiclass import type_of_target
from ..base import _pprint

__all__ = [
    "BaseCrossValidator",
    "KFold",
    "GroupKFold",
    "LeaveOneGroupOut",
    "LeaveOneOut",
    "LeavePGroupsOut",
    "LeavePOut",
    "RepeatedStratifiedKFold",
    "RepeatedKFold",
    "ShuffleSplit",
    "GroupShuffleSplit",
    "StratifiedKFold",
    "StratifiedGroupKFold",
    "StratifiedShuffleSplit",
    "PredefinedSplit",
    "train_test_split",
    "check_cv",
]


class BaseCrossValidator(metaclass=ABCMeta):
    """Base class for all cross-validators

    Implementations must define `_iter_test_masks` or `_iter_test_indices`.
    """

    def split(self, X, y=None, groups=None):
        """Generate indices to split data into training and test set.

        Parameters
        ----------
        X : array-like of shape (n_samples, n_features)
            Training data, where `n_samples` is the number of samples
            and `n_features` is the number of features.

        y : array-like of shape (n_samples,)
            The target variable for supervised learning problems.

        groups : array-like of shape (n_samples,), default=None
            Group labels for the samples used while splitting the dataset into
            train/test set.

        Yields
        ------
        train : ndarray
            The training set indices for that split.

        test : ndarray
            The testing set indices for that split.
        """
        X, y, groups = indexable(X, y, groups)
        indices = np.arange(_num_samples(X))
        for test_index in self._iter_test_masks(X, y, groups):
            train_index = indices[np.logical_not(test_index)]
            test_index = indices[test_index]
            yield train_index, test_index

    # Since subclasses must implement either _iter_test_masks or
    # _iter_test_indices, neither can be abstract.
    def _iter_test_masks(self, X=None, y=None, groups=None):
        """Generates boolean masks corresponding to test sets.

        By default, delegates to _iter_test_indices(X, y, groups)
        """
        for test_index in self._iter_test_indices(X, y, groups):
            test_mask = np.zeros(_num_samples(X), dtype=bool)
            test_mask[test_index] = True
            yield test_mask

    def _iter_test_indices(self, X=None, y=None, groups=None):
        """Generates integer indices corresponding to test sets."""
        raise NotImplementedError

    @abstractmethod
    def get_n_splits(self, X=None, y=None, groups=None):
        """Returns the number of splitting iterations in the cross-validator"""

    def __repr__(self):
        return _build_repr(self)


class LeaveOneOut(BaseCrossValidator):
    """Leave-One-Out cross-validator

    Provides train/test indices to split data in train/test sets. Each
    sample is used once as a test set (singleton) while the remaining
    samples form the training set.

    Note: ``LeaveOneOut()`` is equivalent to ``KFold(n_splits=n)`` and
    ``LeavePOut(p=1)`` where ``n`` is the number of samples.

    Due to the high number of test sets (which is the same as the
    number of samples) this cross-validation method can be very costly.
    For large datasets one should favor :class:`KFold`, :class:`ShuffleSplit`
    or :class:`StratifiedKFold`.

    Read more in the :ref:`User Guide <leave_one_out>`.

    Examples
    --------
    >>> import numpy as np
    >>> from sklearn.model_selection import LeaveOneOut
    >>> X = np.array([[1, 2], [3, 4]])
    >>> y = np.array([1, 2])
    >>> loo = LeaveOneOut()
    >>> loo.get_n_splits(X)
    2
    >>> print(loo)
    LeaveOneOut()
    >>> for train_index, test_index in loo.split(X):
    ...     print("TRAIN:", train_index, "TEST:", test_index)
    ...     X_train, X_test = X[train_index], X[test_index]
    ...     y_train, y_test = y[train_index], y[test_index]
    ...     print(X_train, X_test, y_train, y_test)
    TRAIN: [1] TEST: [0]
    [[3 4]] [[1 2]] [2] [1]
    TRAIN: [0] TEST: [1]
    [[1 2]] [[3 4]] [1] [2]

    See Also
    --------
    LeaveOneGroupOut : For splitting the data according to explicit,
        domain-specific stratification of the dataset.
    GroupKFold : K-fold iterator variant with non-overlapping groups.
    """

    def _iter_test_indices(self, X, y=None, groups=None):
        n_samples = _num_samples(X)
        if n_samples <= 1:
            raise ValueError(
                "Cannot perform LeaveOneOut with n_samples={}.".format(n_samples)
            )
        return range(n_samples)

    def get_n_splits(self, X, y=None, groups=None):
        """Returns the number of splitting iterations in the cross-validator

        Parameters
        ----------
        X : array-like of shape (n_samples, n_features)
            Training data, where `n_samples` is the number of samples
            and `n_features` is the number of features.

        y : object
            Always ignored, exists for compatibility.

        groups : object
            Always ignored, exists for compatibility.

        Returns
        -------
        n_splits : int
            Returns the number of splitting iterations in the cross-validator.
        """
        if X is None:
            raise ValueError("The 'X' parameter should not be None.")
        return _num_samples(X)


class LeavePOut(BaseCrossValidator):
    """Leave-P-Out cross-validator

    Provides train/test indices to split data in train/test sets. This results
    in testing on all distinct samples of size p, while the remaining n - p
    samples form the training set in each iteration.

    Note: ``LeavePOut(p)`` is NOT equivalent to
    ``KFold(n_splits=n_samples // p)`` which creates non-overlapping test sets.

    Due to the high number of iterations which grows combinatorically with the
    number of samples this cross-validation method can be very costly. For
    large datasets one should favor :class:`KFold`, :class:`StratifiedKFold`
    or :class:`ShuffleSplit`.

    Read more in the :ref:`User Guide <leave_p_out>`.

    Parameters
    ----------
    p : int
        Size of the test sets. Must be strictly less than the number of
        samples.

    Examples
    --------
    >>> import numpy as np
    >>> from sklearn.model_selection import LeavePOut
    >>> X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]])
    >>> y = np.array([1, 2, 3, 4])
    >>> lpo = LeavePOut(2)
    >>> lpo.get_n_splits(X)
    6
    >>> print(lpo)
    LeavePOut(p=2)
    >>> for train_index, test_index in lpo.split(X):
    ...     print("TRAIN:", train_index, "TEST:", test_index)
    ...     X_train, X_test = X[train_index], X[test_index]
    ...     y_train, y_test = y[train_index], y[test_index]
    TRAIN: [2 3] TEST: [0 1]
    TRAIN: [1 3] TEST: [0 2]
    TRAIN: [1 2] TEST: [0 3]
    TRAIN: [0 3] TEST: [1 2]
    TRAIN: [0 2] TEST: [1 3]
    TRAIN: [0 1] TEST: [2 3]
    """

    def __init__(self, p):
        self.p = p

    def _iter_test_indices(self, X, y=None, groups=None):
        n_samples = _num_samples(X)
        if n_samples <= self.p:
            raise ValueError(
                "p={} must be strictly less than the number of samples={}".format(
                    self.p, n_samples
                )
            )
        for combination in combinations(range(n_samples), self.p):
            yield np.array(combination)

    def get_n_splits(self, X, y=None, groups=None):
        """Returns the number of splitting iterations in the cross-validator

        Parameters
        ----------
        X : array-like of shape (n_samples, n_features)
            Training data, where `n_samples` is the number of samples
            and `n_features` is the number of features.

        y : object
            Always ignored, exists for compatibility.

        groups : object
            Always ignored, exists for compatibility.
        """
        if X is None:
            raise ValueError("The 'X' parameter should not be None.")
        return int(comb(_num_samples(X), self.p, exact=True))


class _BaseKFold(BaseCrossValidator, metaclass=ABCMeta):
    """Base class for KFold, GroupKFold, and StratifiedKFold"""

    @abstractmethod
    def __init__(self, n_splits, *, shuffle, random_state):
        if not isinstance(n_splits, numbers.Integral):
            raise ValueError(
                "The number of folds must be of Integral type. "
                "%s of type %s was passed." % (n_splits, type(n_splits))
            )
        n_splits = int(n_splits)

        if n_splits <= 1:
            raise ValueError(
                "k-fold cross-validation requires at least one"
                " train/test split by setting n_splits=2 or more,"
                " got n_splits={0}.".format(n_splits)
            )

        if not isinstance(shuffle, bool):
            raise TypeError("shuffle must be True or False; got {0}".format(shuffle))

        if not shuffle and random_state is not None:  # None is the default
            raise ValueError(
                "Setting a random_state has no effect since shuffle is "
                "False. You should leave "
                "random_state to its default (None), or set shuffle=True.",
            )

        self.n_splits = n_splits
        self.shuffle = shuffle
        self.random_state = random_state

    def split(self, X, y=None, groups=None):
        """Generate indices to split data into training and test set.

        Parameters
        ----------
        X : array-like of shape (n_samples, n_features)
            Training data, where `n_samples` is the number of samples
            and `n_features` is the number of features.

        y : array-like of shape (n_samples,), default=None
            The target variable for supervised learning problems.

        groups : array-like of shape (n_samples,), default=None
            Group labels for the samples used while splitting the dataset into
            train/test set.

        Yields
        ------
        train : ndarray
            The training set indices for that split.

        test : ndarray
            The testing set indices for that split.
        """
        X, y, groups = indexable(X, y, groups)
        n_samples = _num_samples(X)
        if self.n_splits > n_samples:
            raise ValueError(
                (
                    "Cannot have number of splits n_splits={0} greater"
                    " than the number of samples: n_samples={1}."
                ).format(self.n_splits, n_samples)
            )

        for train, test in super().split(X, y, groups):
            yield train, test

    def get_n_splits(self, X=None, y=None, groups=None):
        """Returns the number of splitting iterations in the cross-validator

        Parameters
        ----------
        X : object
            Always ignored, exists for compatibility.

        y : object
            Always ignored, exists for compatibility.

        groups : object
            Always ignored, exists for compatibility.

        Returns
        -------
        n_splits : int
            Returns the number of splitting iterations in the cross-validator.
        """
        return self.n_splits


class KFold(_BaseKFold):
    """K-Folds cross-validator

    Provides train/test indices to split data in train/test sets. Split
    dataset into k consecutive folds (without shuffling by default).

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

    Read more in the :ref:`User Guide <k_fold>`.

    Parameters
    ----------
    n_splits : int, default=5
        Number of folds. Must be at least 2.

        .. versionchanged:: 0.22
            ``n_splits`` default value changed from 3 to 5.

    shuffle : bool, default=False
        Whether to shuffle the data before splitting into batches.
        Note that the samples within each split will not be shuffled.

    random_state : int, RandomState instance or None, default=None
        When `shuffle` is True, `random_state` affects the ordering of the
        indices, which controls the randomness of each fold. Otherwise, this
        parameter has no effect.
        Pass an int for reproducible output across multiple function calls.
        See :term:`Glossary <random_state>`.

    Examples
    --------
    >>> import numpy as np
    >>> from sklearn.model_selection import KFold
    >>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
    >>> y = np.array([1, 2, 3, 4])
    >>> kf = KFold(n_splits=2)
    >>> kf.get_n_splits(X)
    2
    >>> print(kf)
    KFold(n_splits=2, random_state=None, shuffle=False)
    >>> for train_index, test_index in kf.split(X):
    ...     print("TRAIN:", train_index, "TEST:", test_index)
    ...     X_train, X_test = X[train_index], X[test_index]
    ...     y_train, y_test = y[train_index], y[test_index]
    TRAIN: [2 3] TEST: [0 1]
    TRAIN: [0 1] TEST: [2 3]

    Notes
    -----
    The first ``n_samples % n_splits`` folds have size
    ``n_samples // n_splits + 1``, other folds have size
    ``n_samples // n_splits``, where ``n_samples`` is the number of samples.

    Randomized CV splitters may return different results for each call of
    split. You can make the results identical by setting `random_state`
    to an integer.

    See Also
    --------
    StratifiedKFold : Takes group information into account to avoid building
        folds with imbalanced class distributions (for binary or multiclass
        classification tasks).

    GroupKFold : K-fold iterator variant with non-overlapping groups.

    RepeatedKFold : Repeats K-Fold n times.
    """

    def __init__(self, n_splits=5, *, shuffle=False, random_state=None):
        super().__init__(n_splits=n_splits, shuffle=shuffle, random_state=random_state)

    def _iter_test_indices(self, X, y=None, groups=None):
        n_samples = _num_samples(X)
        indices = np.arange(n_samples)
        if self.shuffle:
            check_random_state(self.random_state).shuffle(indices)

        n_splits = self.n_splits
        fold_sizes = np.full(n_splits, n_samples // n_splits, dtype=int)
        fold_sizes[: n_samples % n_splits] += 1
        current = 0
        for fold_size in fold_sizes:
            start, stop = current, current + fold_size
            yield indices[start:stop]
            current = stop


class GroupKFold(_BaseKFold):
    """K-fold iterator variant with non-overlapping groups.

    The same group will not appear in two different folds (the number of
    distinct groups has to be at least equal to the number of folds).

    The folds are approximately balanced in the sense that the number of
    distinct groups is approximately the same in each fold.

    Read more in the :ref:`User Guide <group_k_fold>`.

    Parameters
    ----------
    n_splits : int, default=5
        Number of folds. Must be at least 2.

        .. versionchanged:: 0.22
            ``n_splits`` default value changed from 3 to 5.

    Examples
    --------
    >>> import numpy as np
    >>> from sklearn.model_selection import GroupKFold
    >>> X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]])
    >>> y = np.array([1, 2, 3, 4])
    >>> groups = np.array([0, 0, 2, 2])
    >>> group_kfold = GroupKFold(n_splits=2)
    >>> group_kfold.get_n_splits(X, y, groups)
    2
    >>> print(group_kfold)
    GroupKFold(n_splits=2)
    >>> for train_index, test_index in group_kfold.split(X, y, groups):
    ...     print("TRAIN:", train_index, "TEST:", test_index)
    ...     X_train, X_test = X[train_index], X[test_index]
    ...     y_train, y_test = y[train_index], y[test_index]
    ...     print(X_train, X_test, y_train, y_test)
    ...
    TRAIN: [0 1] TEST: [2 3]
    [[1 2]
     [3 4]] [[5 6]
     [7 8]] [1 2] [3 4]
    TRAIN: [2 3] TEST: [0 1]
    [[5 6]
     [7 8]] [[1 2]
     [3 4]] [3 4] [1 2]

    See Also
    --------
    LeaveOneGroupOut : For splitting the data according to explicit
        domain-specific stratification of the dataset.
    """

    def __init__(self, n_splits=5):
        super().__init__(n_splits, shuffle=False, random_state=None)

    def _iter_test_indices(self, X, y, groups):
        if groups is None:
            raise ValueError("The 'groups' parameter should not be None.")
        groups = check_array(groups, ensure_2d=False, dtype=None)

        unique_groups, groups = np.unique(groups, return_inverse=True)
        n_groups = len(unique_groups)

        if self.n_splits > n_groups:
            raise ValueError(
                "Cannot have number of splits n_splits=%d greater"
                " than the number of groups: %d." % (self.n_splits, n_groups)
            )

        # Weight groups by their number of occurrences
        n_samples_per_group = np.bincount(groups)

        # Distribute the most frequent groups first
        indices = np.argsort(n_samples_per_group)[::-1]
        n_samples_per_group = n_samples_per_group[indices]

        # Total weight of each fold
        n_samples_per_fold = np.zeros(self.n_splits)

        # Mapping from group index to fold index
        group_to_fold = np.zeros(len(unique_groups))

        # Distribute samples by adding the largest weight to the lightest fold
        for group_index, weight in enumerate(n_samples_per_group):
            lightest_fold = np.argmin(n_samples_per_fold)
            n_samples_per_fold[lightest_fold] += weight
            group_to_fold[indices[group_index]] = lightest_fold

        indices = group_to_fold[groups]

        for f in range(self.n_splits):
            yield np.where(indices == f)[0]

    def split(self, X, y=None, groups=None):
        """Generate indices to split data into training and test set.

        Parameters
        ----------
        X : array-like of shape (n_samples, n_features)
            Training data, where `n_samples` is the number of samples
            and `n_features` is the number of features.

        y : array-like of shape (n_samples,), default=None
            The target variable for supervised learning problems.

        groups : array-like of shape (n_samples,)
            Group labels for the samples used while splitting the dataset into
            train/test set.

        Yields
        ------
        train : ndarray
            The training set indices for that split.

        test : ndarray
            The testing set indices for that split.
        """
        return super().split(X, y, groups)


class StratifiedKFold(_BaseKFold):
    """Stratified K-Folds cross-validator.

    Provides train/test indices to split data in train/test sets.

    This cross-validation object is a variation of KFold that returns
    stratified folds. The folds are made by preserving the percentage of
    samples for each class.

    Read more in the :ref:`User Guide <stratified_k_fold>`.

    Parameters
    ----------
    n_splits : int, default=5
        Number of folds. Must be at least 2.

        .. versionchanged:: 0.22
            ``n_splits`` default value changed from 3 to 5.

    shuffle : bool, default=False
        Whether to shuffle each class's samples before splitting into batches.
        Note that the samples within each split will not be shuffled.

    random_state : int, RandomState instance or None, default=None
        When `shuffle` is True, `random_state` affects the ordering of the
        indices, which controls the randomness of each fold for each class.
        Otherwise, leave `random_state` as `None`.
        Pass an int for reproducible output across multiple function calls.
        See :term:`Glossary <random_state>`.

    Examples
    --------
    >>> import numpy as np
    >>> from sklearn.model_selection import StratifiedKFold
    >>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
    >>> y = np.array([0, 0, 1, 1])
    >>> skf = StratifiedKFold(n_splits=2)
    >>> skf.get_n_splits(X, y)
    2
    >>> print(skf)
    StratifiedKFold(n_splits=2, random_state=None, shuffle=False)
    >>> for train_index, test_index in skf.split(X, y):
    ...     print("TRAIN:", train_index, "TEST:", test_index)
    ...     X_train, X_test = X[train_index], X[test_index]
    ...     y_train, y_test = y[train_index], y[test_index]
    TRAIN: [1 3] TEST: [0 2]
    TRAIN: [0 2] TEST: [1 3]

    Notes
    -----
    The implementation is designed to:

    * Generate test sets such that all contain the same distribution of
      classes, or as close as possible.
    * Be invariant to class label: relabelling ``y = ["Happy", "Sad"]`` to
      ``y = [1, 0]`` should not change the indices generated.
    * Preserve order dependencies in the dataset ordering, when
      ``shuffle=False``: all samples from class k in some test set were
      contiguous in y, or separated in y by samples from classes other than k.
    * Generate test sets where the smallest and largest differ by at most one
      sample.

    .. versionchanged:: 0.22
        The previous implementation did not follow the last constraint.

    See Also
    --------
    RepeatedStratifiedKFold : Repeats Stratified K-Fold n times.
    """

    def __init__(self, n_splits=5, *, shuffle=False, random_state=None):
        super().__init__(n_splits=n_splits, shuffle=shuffle, random_state=random_state)

    def _make_test_folds(self, X, y=None):
        rng = check_random_state(self.random_state)
        y = np.asarray(y)
        type_of_target_y = type_of_target(y)
        allowed_target_types = ("binary", "multiclass")
        if type_of_target_y not in allowed_target_types:
            raise ValueError(
                "Supported target types are: {}. Got {!r} instead.".format(
                    allowed_target_types, type_of_target_y
                )
            )

        y = column_or_1d(y)

        _, y_idx, y_inv = np.unique(y, return_index=True, return_inverse=True)
        # y_inv encodes y according to lexicographic order. We invert y_idx to
        # map the classes so that they are encoded by order of appearance:
        # 0 represents the first label appearing in y, 1 the second, etc.
        _, class_perm = np.unique(y_idx, return_inverse=True)
        y_encoded = class_perm[y_inv]

        n_classes = len(y_idx)
        y_counts = np.bincount(y_encoded)
        min_groups = np.min(y_counts)
        if np.all(self.n_splits > y_counts):
            raise ValueError(
                "n_splits=%d cannot be greater than the"
                " number of members in each class." % (self.n_splits)
            )
        if self.n_splits > min_groups:
            warnings.warn(
                "The least populated class in y has only %d"
                " members, which is less than n_splits=%d."
                % (min_groups, self.n_splits),
                UserWarning,
            )

        # Determine the optimal number of samples from each class in each fold,
        # using round robin over the sorted y. (This can be done direct from
        # counts, but that code is unreadable.)
        y_order = np.sort(y_encoded)
        allocation = np.asarray(
            [
                np.bincount(y_order[i :: self.n_splits], minlength=n_classes)
                for i in range(self.n_splits)
            ]
        )

        # To maintain the data order dependencies as best as possible within
        # the stratification constraint, we assign samples from each class in
        # blocks (and then mess that up when shuffle=True).
        test_folds = np.empty(len(y), dtype="i")
        for k in range(n_classes):
            # since the kth column of allocation stores the number of samples
            # of class k in each test set, this generates blocks of fold
            # indices corresponding to the allocation for class k.
            folds_for_class = np.arange(self.n_splits).repeat(allocation[:, k])
            if self.shuffle:
                rng.shuffle(folds_for_class)
            test_folds[y_encoded == k] = folds_for_class
        return test_folds

    def _iter_test_masks(self, X, y=None, groups=None):
        test_folds = self._make_test_folds(X, y)
        for i in range(self.n_splits):
            yield test_folds == i

[docs] def split(self, X, y, groups=None): """Generate indices to split data into training and test set. Parameters ---------- X : array-like of shape (n_samples, n_features) Training data, where `n_samples` is the number of samples and `n_features` is the number of features. Note that providing ``y`` is sufficient to generate the splits and hence ``np.zeros(n_samples)`` may be used as a placeholder for ``X`` instead of actual training data. y : array-like of shape (n_samples,) The target variable for supervised learning problems. Stratification is done based on the y labels. groups : object Always ignored, exists for compatibility. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. Notes ----- Randomized CV splitters may return different results for each call of split. You can make the results identical by setting `random_state` to an integer. """ y = check_array(y, ensure_2d=False, dtype=None) return super().split(X, y, groups)
class StratifiedGroupKFold(_BaseKFold): """Stratified K-Folds iterator variant with non-overlapping groups. This cross-validation object is a variation of StratifiedKFold attempts to return stratified folds with non-overlapping groups. The folds are made by preserving the percentage of samples for each class. The same group will not appear in two different folds (the number of distinct groups has to be at least equal to the number of folds). The difference between GroupKFold and StratifiedGroupKFold is that the former attempts to create balanced folds such that the number of distinct groups is approximately the same in each fold, whereas StratifiedGroupKFold attempts to create folds which preserve the percentage of samples for each class as much as possible given the constraint of non-overlapping groups between splits. Read more in the :ref:`User Guide <cross_validation>`. Parameters ---------- n_splits : int, default=5 Number of folds. Must be at least 2. shuffle : bool, default=False Whether to shuffle each class's samples before splitting into batches. Note that the samples within each split will not be shuffled. This implementation can only shuffle groups that have approximately the same y distribution, no global shuffle will be performed. random_state : int or RandomState instance, default=None When `shuffle` is True, `random_state` affects the ordering of the indices, which controls the randomness of each fold for each class. Otherwise, leave `random_state` as `None`. Pass an int for reproducible output across multiple function calls. See :term:`Glossary <random_state>`. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import StratifiedGroupKFold >>> X = np.ones((17, 2)) >>> y = np.array([0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0]) >>> groups = np.array([1, 1, 2, 2, 3, 3, 3, 4, 5, 5, 5, 5, 6, 6, 7, 8, 8]) >>> cv = StratifiedGroupKFold(n_splits=3) >>> for train_idxs, test_idxs in cv.split(X, y, groups): ... print("TRAIN:", groups[train_idxs]) ... print(" ", y[train_idxs]) ... print(" TEST:", groups[test_idxs]) ... print(" ", y[test_idxs]) TRAIN: [1 1 2 2 4 5 5 5 5 8 8] [0 0 1 1 1 0 0 0 0 0 0] TEST: [3 3 3 6 6 7] [1 1 1 0 0 0] TRAIN: [3 3 3 4 5 5 5 5 6 6 7] [1 1 1 1 0 0 0 0 0 0 0] TEST: [1 1 2 2 8 8] [0 0 1 1 0 0] TRAIN: [1 1 2 2 3 3 3 6 6 7 8 8] [0 0 1 1 1 1 1 0 0 0 0 0] TEST: [4 5 5 5 5] [1 0 0 0 0] Notes ----- The implementation is designed to: * Mimic the behavior of StratifiedKFold as much as possible for trivial groups (e.g. when each group contains only one sample). * Be invariant to class label: relabelling ``y = ["Happy", "Sad"]`` to ``y = [1, 0]`` should not change the indices generated. * Stratify based on samples as much as possible while keeping non-overlapping groups constraint. That means that in some cases when there is a small number of groups containing a large number of samples the stratification will not be possible and the behavior will be close to GroupKFold. See also -------- StratifiedKFold: Takes class information into account to build folds which retain class distributions (for binary or multiclass classification tasks). GroupKFold: K-fold iterator variant with non-overlapping groups. """ def __init__(self, n_splits=5, shuffle=False, random_state=None): super().__init__(n_splits=n_splits, shuffle=shuffle, random_state=random_state) def _iter_test_indices(self, X, y, groups): # Implementation is based on this kaggle kernel: # https://www.kaggle.com/jakubwasikowski/stratified-group-k-fold-cross-validation # and is a subject to Apache 2.0 License. You may obtain a copy of the # License at http://www.apache.org/licenses/LICENSE-2.0 # Changelist: # - Refactored function to a class following scikit-learn KFold # interface. # - Added heuristic for assigning group to the least populated fold in # cases when all other criteria are equal # - Swtch from using python ``Counter`` to ``np.unique`` to get class # distribution # - Added scikit-learn checks for input: checking that target is binary # or multiclass, checking passed random state, checking that number # of splits is less than number of members in each class, checking # that least populated class has more members than there are splits. rng = check_random_state(self.random_state) y = np.asarray(y) type_of_target_y = type_of_target(y) allowed_target_types = ("binary", "multiclass") if type_of_target_y not in allowed_target_types: raise ValueError( "Supported target types are: {}. Got {!r} instead.".format( allowed_target_types, type_of_target_y ) ) y = column_or_1d(y) _, y_inv, y_cnt = np.unique(y, return_inverse=True, return_counts=True) if np.all(self.n_splits > y_cnt): raise ValueError( "n_splits=%d cannot be greater than the" " number of members in each class." % (self.n_splits) ) n_smallest_class = np.min(y_cnt) if self.n_splits > n_smallest_class: warnings.warn( "The least populated class in y has only %d" " members, which is less than n_splits=%d." % (n_smallest_class, self.n_splits), UserWarning, ) n_classes = len(y_cnt) _, groups_inv, groups_cnt = np.unique( groups, return_inverse=True, return_counts=True ) y_counts_per_group = np.zeros((len(groups_cnt), n_classes)) for class_idx, group_idx in zip(y_inv, groups_inv): y_counts_per_group[group_idx, class_idx] += 1 y_counts_per_fold = np.zeros((self.n_splits, n_classes)) groups_per_fold = defaultdict(set) if self.shuffle: rng.shuffle(y_counts_per_group) # Stable sort to keep shuffled order for groups with the same # class distribution variance sorted_groups_idx = np.argsort( -np.std(y_counts_per_group, axis=1), kind="mergesort" ) for group_idx in sorted_groups_idx: group_y_counts = y_counts_per_group[group_idx] best_fold = self._find_best_fold( y_counts_per_fold=y_counts_per_fold, y_cnt=y_cnt, group_y_counts=group_y_counts, ) y_counts_per_fold[best_fold] += group_y_counts groups_per_fold[best_fold].add(group_idx) for i in range(self.n_splits): test_indices = [ idx for idx, group_idx in enumerate(groups_inv) if group_idx in groups_per_fold[i] ] yield test_indices def _find_best_fold(self, y_counts_per_fold, y_cnt, group_y_counts): best_fold = None min_eval = np.inf min_samples_in_fold = np.inf for i in range(self.n_splits): y_counts_per_fold[i] += group_y_counts # Summarise the distribution over classes in each proposed fold std_per_class = np.std(y_counts_per_fold / y_cnt.reshape(1, -1), axis=0) y_counts_per_fold[i] -= group_y_counts fold_eval = np.mean(std_per_class) samples_in_fold = np.sum(y_counts_per_fold[i]) is_current_fold_better = ( fold_eval < min_eval or np.isclose(fold_eval, min_eval) and samples_in_fold < min_samples_in_fold ) if is_current_fold_better: min_eval = fold_eval min_samples_in_fold = samples_in_fold best_fold = i return best_fold class TimeSeriesSplit(_BaseKFold): """Time Series cross-validator Provides train/test indices to split time series data samples that are observed at fixed time intervals, in train/test sets. In each split, test indices must be higher than before, and thus shuffling in cross validator is inappropriate. This cross-validation object is a variation of :class:`KFold`. In the kth split, it returns first k folds as train set and the (k+1)th fold as test set. Note that unlike standard cross-validation methods, successive training sets are supersets of those that come before them. Read more in the :ref:`User Guide <time_series_split>`. .. versionadded:: 0.18 Parameters ---------- n_splits : int, default=5 Number of splits. Must be at least 2. .. versionchanged:: 0.22 ``n_splits`` default value changed from 3 to 5. max_train_size : int, default=None Maximum size for a single training set. test_size : int, default=None Used to limit the size of the test set. Defaults to ``n_samples // (n_splits + 1)``, which is the maximum allowed value with ``gap=0``. .. versionadded:: 0.24 gap : int, default=0 Number of samples to exclude from the end of each train set before the test set. .. versionadded:: 0.24 Examples -------- >>> import numpy as np >>> from sklearn.model_selection import TimeSeriesSplit >>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4], [1, 2], [3, 4]]) >>> y = np.array([1, 2, 3, 4, 5, 6]) >>> tscv = TimeSeriesSplit() >>> print(tscv) TimeSeriesSplit(gap=0, max_train_size=None, n_splits=5, test_size=None) >>> for train_index, test_index in tscv.split(X): ... print("TRAIN:", train_index, "TEST:", test_index) ... X_train, X_test = X[train_index], X[test_index] ... y_train, y_test = y[train_index], y[test_index] TRAIN: [0] TEST: [1] TRAIN: [0 1] TEST: [2] TRAIN: [0 1 2] TEST: [3] TRAIN: [0 1 2 3] TEST: [4] TRAIN: [0 1 2 3 4] TEST: [5] >>> # Fix test_size to 2 with 12 samples >>> X = np.random.randn(12, 2) >>> y = np.random.randint(0, 2, 12) >>> tscv = TimeSeriesSplit(n_splits=3, test_size=2) >>> for train_index, test_index in tscv.split(X): ... print("TRAIN:", train_index, "TEST:", test_index) ... X_train, X_test = X[train_index], X[test_index] ... y_train, y_test = y[train_index], y[test_index] TRAIN: [0 1 2 3 4 5] TEST: [6 7] TRAIN: [0 1 2 3 4 5 6 7] TEST: [8 9] TRAIN: [0 1 2 3 4 5 6 7 8 9] TEST: [10 11] >>> # Add in a 2 period gap >>> tscv = TimeSeriesSplit(n_splits=3, test_size=2, gap=2) >>> for train_index, test_index in tscv.split(X): ... print("TRAIN:", train_index, "TEST:", test_index) ... X_train, X_test = X[train_index], X[test_index] ... y_train, y_test = y[train_index], y[test_index] TRAIN: [0 1 2 3] TEST: [6 7] TRAIN: [0 1 2 3 4 5] TEST: [8 9] TRAIN: [0 1 2 3 4 5 6 7] TEST: [10 11] Notes ----- The training set has size ``i * n_samples // (n_splits + 1) + n_samples % (n_splits + 1)`` in the ``i`` th split, with a test set of size ``n_samples//(n_splits + 1)`` by default, where ``n_samples`` is the number of samples. """ def __init__(self, n_splits=5, *, max_train_size=None, test_size=None, gap=0): super().__init__(n_splits, shuffle=False, random_state=None) self.max_train_size = max_train_size self.test_size = test_size self.gap = gap def split(self, X, y=None, groups=None): """Generate indices to split data into training and test set. Parameters ---------- X : array-like of shape (n_samples, n_features) Training data, where `n_samples` is the number of samples and `n_features` is the number of features. y : array-like of shape (n_samples,) Always ignored, exists for compatibility. groups : array-like of shape (n_samples,) Always ignored, exists for compatibility. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. """ X, y, groups = indexable(X, y, groups) n_samples = _num_samples(X) n_splits = self.n_splits n_folds = n_splits + 1 gap = self.gap test_size = ( self.test_size if self.test_size is not None else n_samples // n_folds ) # Make sure we have enough samples for the given split parameters if n_folds > n_samples: raise ValueError( f"Cannot have number of folds={n_folds} greater" f" than the number of samples={n_samples}." ) if n_samples - gap - (test_size * n_splits) <= 0: raise ValueError( f"Too many splits={n_splits} for number of samples" f"={n_samples} with test_size={test_size} and gap={gap}." ) indices = np.arange(n_samples) test_starts = range(n_samples - n_splits * test_size, n_samples, test_size) for test_start in test_starts: train_end = test_start - gap if self.max_train_size and self.max_train_size < train_end: yield ( indices[train_end - self.max_train_size : train_end], indices[test_start : test_start + test_size], ) else: yield ( indices[:train_end], indices[test_start : test_start + test_size], ) class LeaveOneGroupOut(BaseCrossValidator): """Leave One Group Out cross-validator Provides train/test indices to split data according to a third-party provided group. This group information can be used to encode arbitrary domain specific stratifications of the samples as integers. For instance the groups could be the year of collection of the samples and thus allow for cross-validation against time-based splits. Read more in the :ref:`User Guide <leave_one_group_out>`. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import LeaveOneGroupOut >>> X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]]) >>> y = np.array([1, 2, 1, 2]) >>> groups = np.array([1, 1, 2, 2]) >>> logo = LeaveOneGroupOut() >>> logo.get_n_splits(X, y, groups) 2 >>> logo.get_n_splits(groups=groups) # 'groups' is always required 2 >>> print(logo) LeaveOneGroupOut() >>> for train_index, test_index in logo.split(X, y, groups): ... print("TRAIN:", train_index, "TEST:", test_index) ... X_train, X_test = X[train_index], X[test_index] ... y_train, y_test = y[train_index], y[test_index] ... print(X_train, X_test, y_train, y_test) TRAIN: [2 3] TEST: [0 1] [[5 6] [7 8]] [[1 2] [3 4]] [1 2] [1 2] TRAIN: [0 1] TEST: [2 3] [[1 2] [3 4]] [[5 6] [7 8]] [1 2] [1 2] """ def _iter_test_masks(self, X, y, groups): if groups is None: raise ValueError("The 'groups' parameter should not be None.") # We make a copy of groups to avoid side-effects during iteration groups = check_array(groups, copy=True, ensure_2d=False, dtype=None) unique_groups = np.unique(groups) if len(unique_groups) <= 1: raise ValueError( "The groups parameter contains fewer than 2 unique groups " "(%s). LeaveOneGroupOut expects at least 2." % unique_groups ) for i in unique_groups: yield groups == i def get_n_splits(self, X=None, y=None, groups=None): """Returns the number of splitting iterations in the cross-validator Parameters ---------- X : object Always ignored, exists for compatibility. y : object Always ignored, exists for compatibility. groups : array-like of shape (n_samples,) Group labels for the samples used while splitting the dataset into train/test set. This 'groups' parameter must always be specified to calculate the number of splits, though the other parameters can be omitted. Returns ------- n_splits : int Returns the number of splitting iterations in the cross-validator. """ if groups is None: raise ValueError("The 'groups' parameter should not be None.") groups = check_array(groups, ensure_2d=False, dtype=None) return len(np.unique(groups)) def split(self, X, y=None, groups=None): """Generate indices to split data into training and test set. Parameters ---------- X : array-like of shape (n_samples, n_features) Training data, where `n_samples` is the number of samples and `n_features` is the number of features. y : array-like of shape (n_samples,), default=None The target variable for supervised learning problems. groups : array-like of shape (n_samples,) Group labels for the samples used while splitting the dataset into train/test set. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. """ return super().split(X, y, groups) class LeavePGroupsOut(BaseCrossValidator): """Leave P Group(s) Out cross-validator Provides train/test indices to split data according to a third-party provided group. This group information can be used to encode arbitrary domain specific stratifications of the samples as integers. For instance the groups could be the year of collection of the samples and thus allow for cross-validation against time-based splits. The difference between LeavePGroupsOut and LeaveOneGroupOut is that the former builds the test sets with all the samples assigned to ``p`` different values of the groups while the latter uses samples all assigned the same groups. Read more in the :ref:`User Guide <leave_p_groups_out>`. Parameters ---------- n_groups : int Number of groups (``p``) to leave out in the test split. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import LeavePGroupsOut >>> X = np.array([[1, 2], [3, 4], [5, 6]]) >>> y = np.array([1, 2, 1]) >>> groups = np.array([1, 2, 3]) >>> lpgo = LeavePGroupsOut(n_groups=2) >>> lpgo.get_n_splits(X, y, groups) 3 >>> lpgo.get_n_splits(groups=groups) # 'groups' is always required 3 >>> print(lpgo) LeavePGroupsOut(n_groups=2) >>> for train_index, test_index in lpgo.split(X, y, groups): ... print("TRAIN:", train_index, "TEST:", test_index) ... X_train, X_test = X[train_index], X[test_index] ... y_train, y_test = y[train_index], y[test_index] ... print(X_train, X_test, y_train, y_test) TRAIN: [2] TEST: [0 1] [[5 6]] [[1 2] [3 4]] [1] [1 2] TRAIN: [1] TEST: [0 2] [[3 4]] [[1 2] [5 6]] [2] [1 1] TRAIN: [0] TEST: [1 2] [[1 2]] [[3 4] [5 6]] [1] [2 1] See Also -------- GroupKFold : K-fold iterator variant with non-overlapping groups. """ def __init__(self, n_groups): self.n_groups = n_groups def _iter_test_masks(self, X, y, groups): if groups is None: raise ValueError("The 'groups' parameter should not be None.") groups = check_array(groups, copy=True, ensure_2d=False, dtype=None) unique_groups = np.unique(groups) if self.n_groups >= len(unique_groups): raise ValueError( "The groups parameter contains fewer than (or equal to) " "n_groups (%d) numbers of unique groups (%s). LeavePGroupsOut " "expects that at least n_groups + 1 (%d) unique groups be " "present" % (self.n_groups, unique_groups, self.n_groups + 1) ) combi = combinations(range(len(unique_groups)), self.n_groups) for indices in combi: test_index = np.zeros(_num_samples(X), dtype=bool) for l in unique_groups[np.array(indices)]: test_index[groups == l] = True yield test_index def get_n_splits(self, X=None, y=None, groups=None): """Returns the number of splitting iterations in the cross-validator Parameters ---------- X : object Always ignored, exists for compatibility. y : object Always ignored, exists for compatibility. groups : array-like of shape (n_samples,) Group labels for the samples used while splitting the dataset into train/test set. This 'groups' parameter must always be specified to calculate the number of splits, though the other parameters can be omitted. Returns ------- n_splits : int Returns the number of splitting iterations in the cross-validator. """ if groups is None: raise ValueError("The 'groups' parameter should not be None.") groups = check_array(groups, ensure_2d=False, dtype=None) return int(comb(len(np.unique(groups)), self.n_groups, exact=True)) def split(self, X, y=None, groups=None): """Generate indices to split data into training and test set. Parameters ---------- X : array-like of shape (n_samples, n_features) Training data, where `n_samples` is the number of samples and `n_features` is the number of features. y : array-like of shape (n_samples,), default=None The target variable for supervised learning problems. groups : array-like of shape (n_samples,) Group labels for the samples used while splitting the dataset into train/test set. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. """ return super().split(X, y, groups) class _RepeatedSplits(metaclass=ABCMeta): """Repeated splits for an arbitrary randomized CV splitter. Repeats splits for cross-validators n times with different randomization in each repetition. Parameters ---------- cv : callable Cross-validator class. n_repeats : int, default=10 Number of times cross-validator needs to be repeated. random_state : int, RandomState instance or None, default=None Passes `random_state` to the arbitrary repeating cross validator. Pass an int for reproducible output across multiple function calls. See :term:`Glossary <random_state>`. **cvargs : additional params Constructor parameters for cv. Must not contain random_state and shuffle. """ def __init__(self, cv, *, n_repeats=10, random_state=None, **cvargs): if not isinstance(n_repeats, numbers.Integral): raise ValueError("Number of repetitions must be of Integral type.") if n_repeats <= 0: raise ValueError("Number of repetitions must be greater than 0.") if any(key in cvargs for key in ("random_state", "shuffle")): raise ValueError("cvargs must not contain random_state or shuffle.") self.cv = cv self.n_repeats = n_repeats self.random_state = random_state self.cvargs = cvargs def split(self, X, y=None, groups=None): """Generates indices to split data into training and test set. Parameters ---------- X : array-like of shape (n_samples, n_features) Training data, where `n_samples` is the number of samples and `n_features` is the number of features. y : array-like of shape (n_samples,) The target variable for supervised learning problems. groups : array-like of shape (n_samples,), default=None Group labels for the samples used while splitting the dataset into train/test set. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. """ n_repeats = self.n_repeats rng = check_random_state(self.random_state) for idx in range(n_repeats): cv = self.cv(random_state=rng, shuffle=True, **self.cvargs) for train_index, test_index in cv.split(X, y, groups): yield train_index, test_index def get_n_splits(self, X=None, y=None, groups=None): """Returns the number of splitting iterations in the cross-validator Parameters ---------- X : object Always ignored, exists for compatibility. ``np.zeros(n_samples)`` may be used as a placeholder. y : object Always ignored, exists for compatibility. ``np.zeros(n_samples)`` may be used as a placeholder. groups : array-like of shape (n_samples,), default=None Group labels for the samples used while splitting the dataset into train/test set. Returns ------- n_splits : int Returns the number of splitting iterations in the cross-validator. """ rng = check_random_state(self.random_state) cv = self.cv(random_state=rng, shuffle=True, **self.cvargs) return cv.get_n_splits(X, y, groups) * self.n_repeats def __repr__(self): return _build_repr(self) class RepeatedKFold(_RepeatedSplits): """Repeated K-Fold cross validator. Repeats K-Fold n times with different randomization in each repetition. Read more in the :ref:`User Guide <repeated_k_fold>`. Parameters ---------- n_splits : int, default=5 Number of folds. Must be at least 2. n_repeats : int, default=10 Number of times cross-validator needs to be repeated. random_state : int, RandomState instance or None, default=None Controls the randomness of each repeated cross-validation instance. Pass an int for reproducible output across multiple function calls. See :term:`Glossary <random_state>`. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import RepeatedKFold >>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]]) >>> y = np.array([0, 0, 1, 1]) >>> rkf = RepeatedKFold(n_splits=2, n_repeats=2, random_state=2652124) >>> for train_index, test_index in rkf.split(X): ... print("TRAIN:", train_index, "TEST:", test_index) ... X_train, X_test = X[train_index], X[test_index] ... y_train, y_test = y[train_index], y[test_index] ... TRAIN: [0 1] TEST: [2 3] TRAIN: [2 3] TEST: [0 1] TRAIN: [1 2] TEST: [0 3] TRAIN: [0 3] TEST: [1 2] Notes ----- Randomized CV splitters may return different results for each call of split. You can make the results identical by setting `random_state` to an integer. See Also -------- RepeatedStratifiedKFold : Repeats Stratified K-Fold n times. """ def __init__(self, *, n_splits=5, n_repeats=10, random_state=None): super().__init__( KFold, n_repeats=n_repeats, random_state=random_state, n_splits=n_splits ) class RepeatedStratifiedKFold(_RepeatedSplits): """Repeated Stratified K-Fold cross validator. Repeats Stratified K-Fold n times with different randomization in each repetition. Read more in the :ref:`User Guide <repeated_k_fold>`. Parameters ---------- n_splits : int, default=5 Number of folds. Must be at least 2. n_repeats : int, default=10 Number of times cross-validator needs to be repeated. random_state : int, RandomState instance or None, default=None Controls the generation of the random states for each repetition. Pass an int for reproducible output across multiple function calls. See :term:`Glossary <random_state>`. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import RepeatedStratifiedKFold >>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]]) >>> y = np.array([0, 0, 1, 1]) >>> rskf = RepeatedStratifiedKFold(n_splits=2, n_repeats=2, ... random_state=36851234) >>> for train_index, test_index in rskf.split(X, y): ... print("TRAIN:", train_index, "TEST:", test_index) ... X_train, X_test = X[train_index], X[test_index] ... y_train, y_test = y[train_index], y[test_index] ... TRAIN: [1 2] TEST: [0 3] TRAIN: [0 3] TEST: [1 2] TRAIN: [1 3] TEST: [0 2] TRAIN: [0 2] TEST: [1 3] Notes ----- Randomized CV splitters may return different results for each call of split. You can make the results identical by setting `random_state` to an integer. See Also -------- RepeatedKFold : Repeats K-Fold n times. """ def __init__(self, *, n_splits=5, n_repeats=10, random_state=None): super().__init__( StratifiedKFold, n_repeats=n_repeats, random_state=random_state, n_splits=n_splits, ) class BaseShuffleSplit(metaclass=ABCMeta): """Base class for ShuffleSplit and StratifiedShuffleSplit""" def __init__( self, n_splits=10, *, test_size=None, train_size=None, random_state=None ): self.n_splits = n_splits self.test_size = test_size self.train_size = train_size self.random_state = random_state self._default_test_size = 0.1 def split(self, X, y=None, groups=None): """Generate indices to split data into training and test set. Parameters ---------- X : array-like of shape (n_samples, n_features) Training data, where `n_samples` is the number of samples and `n_features` is the number of features. y : array-like of shape (n_samples,) The target variable for supervised learning problems. groups : array-like of shape (n_samples,), default=None Group labels for the samples used while splitting the dataset into train/test set. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. Notes ----- Randomized CV splitters may return different results for each call of split. You can make the results identical by setting `random_state` to an integer. """ X, y, groups = indexable(X, y, groups) for train, test in self._iter_indices(X, y, groups): yield train, test @abstractmethod def _iter_indices(self, X, y=None, groups=None): """Generate (train, test) indices""" def get_n_splits(self, X=None, y=None, groups=None): """Returns the number of splitting iterations in the cross-validator Parameters ---------- X : object Always ignored, exists for compatibility. y : object Always ignored, exists for compatibility. groups : object Always ignored, exists for compatibility. Returns ------- n_splits : int Returns the number of splitting iterations in the cross-validator. """ return self.n_splits def __repr__(self): return _build_repr(self) class ShuffleSplit(BaseShuffleSplit): """Random permutation cross-validator Yields indices to split data into training and test sets. Note: contrary to other cross-validation strategies, random splits do not guarantee that all folds will be different, although this is still very likely for sizeable datasets. Read more in the :ref:`User Guide <ShuffleSplit>`. Parameters ---------- n_splits : int, default=10 Number of re-shuffling & splitting iterations. test_size : float or int, default=None If float, should be between 0.0 and 1.0 and represent the proportion of the dataset to include in the test split. If int, represents the absolute number of test samples. If None, the value is set to the complement of the train size. If ``train_size`` is also None, it will be set to 0.1. train_size : float or int, default=None If float, should be between 0.0 and 1.0 and represent the proportion of the dataset to include in the train split. If int, represents the absolute number of train samples. If None, the value is automatically set to the complement of the test size. random_state : int, RandomState instance or None, default=None Controls the randomness of the training and testing indices produced. Pass an int for reproducible output across multiple function calls. See :term:`Glossary <random_state>`. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import ShuffleSplit >>> X = np.array([[1, 2], [3, 4], [5, 6], [7, 8], [3, 4], [5, 6]]) >>> y = np.array([1, 2, 1, 2, 1, 2]) >>> rs = ShuffleSplit(n_splits=5, test_size=.25, random_state=0) >>> rs.get_n_splits(X) 5 >>> print(rs) ShuffleSplit(n_splits=5, random_state=0, test_size=0.25, train_size=None) >>> for train_index, test_index in rs.split(X): ... print("TRAIN:", train_index, "TEST:", test_index) TRAIN: [1 3 0 4] TEST: [5 2] TRAIN: [4 0 2 5] TEST: [1 3] TRAIN: [1 2 4 0] TEST: [3 5] TRAIN: [3 4 1 0] TEST: [5 2] TRAIN: [3 5 1 0] TEST: [2 4] >>> rs = ShuffleSplit(n_splits=5, train_size=0.5, test_size=.25, ... random_state=0) >>> for train_index, test_index in rs.split(X): ... print("TRAIN:", train_index, "TEST:", test_index) TRAIN: [1 3 0] TEST: [5 2] TRAIN: [4 0 2] TEST: [1 3] TRAIN: [1 2 4] TEST: [3 5] TRAIN: [3 4 1] TEST: [5 2] TRAIN: [3 5 1] TEST: [2 4] """ def __init__( self, n_splits=10, *, test_size=None, train_size=None, random_state=None ): super().__init__( n_splits=n_splits, test_size=test_size, train_size=train_size, random_state=random_state, ) self._default_test_size = 0.1 def _iter_indices(self, X, y=None, groups=None): n_samples = _num_samples(X) n_train, n_test = _validate_shuffle_split( n_samples, self.test_size, self.train_size, default_test_size=self._default_test_size, ) rng = check_random_state(self.random_state) for i in range(self.n_splits): # random partition permutation = rng.permutation(n_samples) ind_test = permutation[:n_test] ind_train = permutation[n_test : (n_test + n_train)] yield ind_train, ind_test class GroupShuffleSplit(ShuffleSplit): """Shuffle-Group(s)-Out cross-validation iterator Provides randomized train/test indices to split data according to a third-party provided group. This group information can be used to encode arbitrary domain specific stratifications of the samples as integers. For instance the groups could be the year of collection of the samples and thus allow for cross-validation against time-based splits. The difference between LeavePGroupsOut and GroupShuffleSplit is that the former generates splits using all subsets of size ``p`` unique groups, whereas GroupShuffleSplit generates a user-determined number of random test splits, each with a user-determined fraction of unique groups. For example, a less computationally intensive alternative to ``LeavePGroupsOut(p=10)`` would be ``GroupShuffleSplit(test_size=10, n_splits=100)``. Note: The parameters ``test_size`` and ``train_size`` refer to groups, and not to samples, as in ShuffleSplit. Read more in the :ref:`User Guide <group_shuffle_split>`. Parameters ---------- n_splits : int, default=5 Number of re-shuffling & splitting iterations. test_size : float, int, default=0.2 If float, should be between 0.0 and 1.0 and represent the proportion of groups to include in the test split (rounded up). If int, represents the absolute number of test groups. If None, the value is set to the complement of the train size. The default will change in version 0.21. It will remain 0.2 only if ``train_size`` is unspecified, otherwise it will complement the specified ``train_size``. train_size : float or int, default=None If float, should be between 0.0 and 1.0 and represent the proportion of the groups to include in the train split. If int, represents the absolute number of train groups. If None, the value is automatically set to the complement of the test size. random_state : int, RandomState instance or None, default=None Controls the randomness of the training and testing indices produced. Pass an int for reproducible output across multiple function calls. See :term:`Glossary <random_state>`. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import GroupShuffleSplit >>> X = np.ones(shape=(8, 2)) >>> y = np.ones(shape=(8, 1)) >>> groups = np.array([1, 1, 2, 2, 2, 3, 3, 3]) >>> print(groups.shape) (8,) >>> gss = GroupShuffleSplit(n_splits=2, train_size=.7, random_state=42) >>> gss.get_n_splits() 2 >>> for train_idx, test_idx in gss.split(X, y, groups): ... print("TRAIN:", train_idx, "TEST:", test_idx) TRAIN: [2 3 4 5 6 7] TEST: [0 1] TRAIN: [0 1 5 6 7] TEST: [2 3 4] """ def __init__( self, n_splits=5, *, test_size=None, train_size=None, random_state=None ): super().__init__( n_splits=n_splits, test_size=test_size, train_size=train_size, random_state=random_state, ) self._default_test_size = 0.2 def _iter_indices(self, X, y, groups): if groups is None: raise ValueError("The 'groups' parameter should not be None.") groups = check_array(groups, ensure_2d=False, dtype=None) classes, group_indices = np.unique(groups, return_inverse=True) for group_train, group_test in super()._iter_indices(X=classes): # these are the indices of classes in the partition # invert them into data indices train = np.flatnonzero(np.in1d(group_indices, group_train)) test = np.flatnonzero(np.in1d(group_indices, group_test)) yield train, test def split(self, X, y=None, groups=None): """Generate indices to split data into training and test set. Parameters ---------- X : array-like of shape (n_samples, n_features) Training data, where `n_samples` is the number of samples and `n_features` is the number of features. y : array-like of shape (n_samples,), default=None The target variable for supervised learning problems. groups : array-like of shape (n_samples,) Group labels for the samples used while splitting the dataset into train/test set. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. Notes ----- Randomized CV splitters may return different results for each call of split. You can make the results identical by setting `random_state` to an integer. """ return super().split(X, y, groups) class StratifiedShuffleSplit(BaseShuffleSplit): """Stratified ShuffleSplit cross-validator Provides train/test indices to split data in train/test sets. This cross-validation object is a merge of StratifiedKFold and ShuffleSplit, which returns stratified randomized folds. The folds are made by preserving the percentage of samples for each class. Note: like the ShuffleSplit strategy, stratified random splits do not guarantee that all folds will be different, although this is still very likely for sizeable datasets. Read more in the :ref:`User Guide <stratified_shuffle_split>`. Parameters ---------- n_splits : int, default=10 Number of re-shuffling & splitting iterations. test_size : float or int, default=None If float, should be between 0.0 and 1.0 and represent the proportion of the dataset to include in the test split. If int, represents the absolute number of test samples. If None, the value is set to the complement of the train size. If ``train_size`` is also None, it will be set to 0.1. train_size : float or int, default=None If float, should be between 0.0 and 1.0 and represent the proportion of the dataset to include in the train split. If int, represents the absolute number of train samples. If None, the value is automatically set to the complement of the test size. random_state : int, RandomState instance or None, default=None Controls the randomness of the training and testing indices produced. Pass an int for reproducible output across multiple function calls. See :term:`Glossary <random_state>`. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import StratifiedShuffleSplit >>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4], [1, 2], [3, 4]]) >>> y = np.array([0, 0, 0, 1, 1, 1]) >>> sss = StratifiedShuffleSplit(n_splits=5, test_size=0.5, random_state=0) >>> sss.get_n_splits(X, y) 5 >>> print(sss) StratifiedShuffleSplit(n_splits=5, random_state=0, ...) >>> for train_index, test_index in sss.split(X, y): ... print("TRAIN:", train_index, "TEST:", test_index) ... X_train, X_test = X[train_index], X[test_index] ... y_train, y_test = y[train_index], y[test_index] TRAIN: [5 2 3] TEST: [4 1 0] TRAIN: [5 1 4] TEST: [0 2 3] TRAIN: [5 0 2] TEST: [4 3 1] TRAIN: [4 1 0] TEST: [2 3 5] TRAIN: [0 5 1] TEST: [3 4 2] """ def __init__( self, n_splits=10, *, test_size=None, train_size=None, random_state=None ): super().__init__( n_splits=n_splits, test_size=test_size, train_size=train_size, random_state=random_state, ) self._default_test_size = 0.1 def _iter_indices(self, X, y, groups=None): n_samples = _num_samples(X) y = check_array(y, ensure_2d=False, dtype=None) n_train, n_test = _validate_shuffle_split( n_samples, self.test_size, self.train_size, default_test_size=self._default_test_size, ) if y.ndim == 2: # for multi-label y, map each distinct row to a string repr # using join because str(row) uses an ellipsis if len(row) > 1000 y = np.array([" ".join(row.astype("str")) for row in y]) classes, y_indices = np.unique(y, return_inverse=True) n_classes = classes.shape[0] class_counts = np.bincount(y_indices) if np.min(class_counts) < 2: raise ValueError( "The least populated class in y has only 1" " member, which is too few. The minimum" " number of groups for any class cannot" " be less than 2." ) if n_train < n_classes: raise ValueError( "The train_size = %d should be greater or " "equal to the number of classes = %d" % (n_train, n_classes) ) if n_test < n_classes: raise ValueError( "The test_size = %d should be greater or " "equal to the number of classes = %d" % (n_test, n_classes) ) # Find the sorted list of instances for each class: # (np.unique above performs a sort, so code is O(n logn) already) class_indices = np.split( np.argsort(y_indices, kind="mergesort"), np.cumsum(class_counts)[:-1] ) rng = check_random_state(self.random_state) for _ in range(self.n_splits): # if there are ties in the class-counts, we want # to make sure to break them anew in each iteration n_i = _approximate_mode(class_counts, n_train, rng) class_counts_remaining = class_counts - n_i t_i = _approximate_mode(class_counts_remaining, n_test, rng) train = [] test = [] for i in range(n_classes): permutation = rng.permutation(class_counts[i]) perm_indices_class_i = class_indices[i].take(permutation, mode="clip") train.extend(perm_indices_class_i[: n_i[i]]) test.extend(perm_indices_class_i[n_i[i] : n_i[i] + t_i[i]]) train = rng.permutation(train) test = rng.permutation(test) yield train, test def split(self, X, y, groups=None): """Generate indices to split data into training and test set. Parameters ---------- X : array-like of shape (n_samples, n_features) Training data, where `n_samples` is the number of samples and `n_features` is the number of features. Note that providing ``y`` is sufficient to generate the splits and hence ``np.zeros(n_samples)`` may be used as a placeholder for ``X`` instead of actual training data. y : array-like of shape (n_samples,) or (n_samples, n_labels) The target variable for supervised learning problems. Stratification is done based on the y labels. groups : object Always ignored, exists for compatibility. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. Notes ----- Randomized CV splitters may return different results for each call of split. You can make the results identical by setting `random_state` to an integer. """ y = check_array(y, ensure_2d=False, dtype=None) return super().split(X, y, groups) def _validate_shuffle_split(n_samples, test_size, train_size, default_test_size=None): """ Validation helper to check if the test/test sizes are meaningful wrt to the size of the data (n_samples) """ if test_size is None and train_size is None: test_size = default_test_size test_size_type = np.asarray(test_size).dtype.kind train_size_type = np.asarray(train_size).dtype.kind if ( test_size_type == "i" and (test_size >= n_samples or test_size <= 0) or test_size_type == "f" and (test_size <= 0 or test_size >= 1) ): raise ValueError( "test_size={0} should be either positive and smaller" " than the number of samples {1} or a float in the " "(0, 1) range".format(test_size, n_samples) ) if ( train_size_type == "i" and (train_size >= n_samples or train_size <= 0) or train_size_type == "f" and (train_size <= 0 or train_size >= 1) ): raise ValueError( "train_size={0} should be either positive and smaller" " than the number of samples {1} or a float in the " "(0, 1) range".format(train_size, n_samples) ) if train_size is not None and train_size_type not in ("i", "f"): raise ValueError("Invalid value for train_size: {}".format(train_size)) if test_size is not None and test_size_type not in ("i", "f"): raise ValueError("Invalid value for test_size: {}".format(test_size)) if train_size_type == "f" and test_size_type == "f" and train_size + test_size > 1: raise ValueError( "The sum of test_size and train_size = {}, should be in the (0, 1)" " range. Reduce test_size and/or train_size.".format(train_size + test_size) ) if test_size_type == "f": n_test = ceil(test_size * n_samples) elif test_size_type == "i": n_test = float(test_size) if train_size_type == "f": n_train = floor(train_size * n_samples) elif train_size_type == "i": n_train = float(train_size) if train_size is None: n_train = n_samples - n_test elif test_size is None: n_test = n_samples - n_train if n_train + n_test > n_samples: raise ValueError( "The sum of train_size and test_size = %d, " "should be smaller than the number of " "samples %d. Reduce test_size and/or " "train_size." % (n_train + n_test, n_samples) ) n_train, n_test = int(n_train), int(n_test) if n_train == 0: raise ValueError( "With n_samples={}, test_size={} and train_size={}, the " "resulting train set will be empty. Adjust any of the " "aforementioned parameters.".format(n_samples, test_size, train_size) ) return n_train, n_test class PredefinedSplit(BaseCrossValidator): """Predefined split cross-validator Provides train/test indices to split data into train/test sets using a predefined scheme specified by the user with the ``test_fold`` parameter. Read more in the :ref:`User Guide <predefined_split>`. .. versionadded:: 0.16 Parameters ---------- test_fold : array-like of shape (n_samples,) The entry ``test_fold[i]`` represents the index of the test set that sample ``i`` belongs to. It is possible to exclude sample ``i`` from any test set (i.e. include sample ``i`` in every training set) by setting ``test_fold[i]`` equal to -1. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import PredefinedSplit >>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]]) >>> y = np.array([0, 0, 1, 1]) >>> test_fold = [0, 1, -1, 1] >>> ps = PredefinedSplit(test_fold) >>> ps.get_n_splits() 2 >>> print(ps) PredefinedSplit(test_fold=array([ 0, 1, -1, 1])) >>> for train_index, test_index in ps.split(): ... print("TRAIN:", train_index, "TEST:", test_index) ... X_train, X_test = X[train_index], X[test_index] ... y_train, y_test = y[train_index], y[test_index] TRAIN: [1 2 3] TEST: [0] TRAIN: [0 2] TEST: [1 3] """ def __init__(self, test_fold): self.test_fold = np.array(test_fold, dtype=int) self.test_fold = column_or_1d(self.test_fold) self.unique_folds = np.unique(self.test_fold) self.unique_folds = self.unique_folds[self.unique_folds != -1] def split(self, X=None, y=None, groups=None): """Generate indices to split data into training and test set. Parameters ---------- X : object Always ignored, exists for compatibility. y : object Always ignored, exists for compatibility. groups : object Always ignored, exists for compatibility. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. """ ind = np.arange(len(self.test_fold)) for test_index in self._iter_test_masks(): train_index = ind[np.logical_not(test_index)] test_index = ind[test_index] yield train_index, test_index def _iter_test_masks(self): """Generates boolean masks corresponding to test sets.""" for f in self.unique_folds: test_index = np.where(self.test_fold == f)[0] test_mask = np.zeros(len(self.test_fold), dtype=bool) test_mask[test_index] = True yield test_mask def get_n_splits(self, X=None, y=None, groups=None): """Returns the number of splitting iterations in the cross-validator Parameters ---------- X : object Always ignored, exists for compatibility. y : object Always ignored, exists for compatibility. groups : object Always ignored, exists for compatibility. Returns ------- n_splits : int Returns the number of splitting iterations in the cross-validator. """ return len(self.unique_folds) class _CVIterableWrapper(BaseCrossValidator): """Wrapper class for old style cv objects and iterables.""" def __init__(self, cv): self.cv = list(cv) def get_n_splits(self, X=None, y=None, groups=None): """Returns the number of splitting iterations in the cross-validator Parameters ---------- X : object Always ignored, exists for compatibility. y : object Always ignored, exists for compatibility. groups : object Always ignored, exists for compatibility. Returns ------- n_splits : int Returns the number of splitting iterations in the cross-validator. """ return len(self.cv) def split(self, X=None, y=None, groups=None): """Generate indices to split data into training and test set. Parameters ---------- X : object Always ignored, exists for compatibility. y : object Always ignored, exists for compatibility. groups : object Always ignored, exists for compatibility. Yields ------ train : ndarray The training set indices for that split. test : ndarray The testing set indices for that split. """ for train, test in self.cv: yield train, test def check_cv(cv=5, y=None, *, classifier=False): """Input checker utility for building a cross-validator Parameters ---------- cv : int, cross-validation generator or an iterable, default=None Determines the cross-validation splitting strategy. Possible inputs for cv are: - None, to use the default 5-fold cross validation, - integer, to specify the number of folds. - :term:`CV splitter`, - An iterable yielding (train, test) splits as arrays of indices. For integer/None inputs, if classifier is True and ``y`` is either binary or multiclass, :class:`StratifiedKFold` is used. In all other cases, :class:`KFold` is used. Refer :ref:`User Guide <cross_validation>` for the various cross-validation strategies that can be used here. .. versionchanged:: 0.22 ``cv`` default value changed from 3-fold to 5-fold. y : array-like, default=None The target variable for supervised learning problems. classifier : bool, default=False Whether the task is a classification task, in which case stratified KFold will be used. Returns ------- checked_cv : a cross-validator instance. The return value is a cross-validator which generates the train/test splits via the ``split`` method. """ cv = 5 if cv is None else cv if isinstance(cv, numbers.Integral): if ( classifier and (y is not None) and (type_of_target(y) in ("binary", "multiclass")) ): return StratifiedKFold(cv) else: return KFold(cv) if not hasattr(cv, "split") or isinstance(cv, str): if not isinstance(cv, Iterable) or isinstance(cv, str): raise ValueError( "Expected cv as an integer, cross-validation " "object (from sklearn.model_selection) " "or an iterable. Got %s." % cv ) return _CVIterableWrapper(cv) return cv # New style cv objects are passed without any modification def train_test_split( *arrays, test_size=None, train_size=None, random_state=None, shuffle=True, stratify=None, ): """Split arrays or matrices into random train and test subsets. Quick utility that wraps input validation and ``next(ShuffleSplit().split(X, y))`` and application to input data into a single call for splitting (and optionally subsampling) data in a oneliner. Read more in the :ref:`User Guide <cross_validation>`. Parameters ---------- *arrays : sequence of indexables with same length / shape[0] Allowed inputs are lists, numpy arrays, scipy-sparse matrices or pandas dataframes. test_size : float or int, default=None If float, should be between 0.0 and 1.0 and represent the proportion of the dataset to include in the test split. If int, represents the absolute number of test samples. If None, the value is set to the complement of the train size. If ``train_size`` is also None, it will be set to 0.25. train_size : float or int, default=None If float, should be between 0.0 and 1.0 and represent the proportion of the dataset to include in the train split. If int, represents the absolute number of train samples. If None, the value is automatically set to the complement of the test size. random_state : int, RandomState instance or None, default=None Controls the shuffling applied to the data before applying the split. Pass an int for reproducible output across multiple function calls. See :term:`Glossary <random_state>`. shuffle : bool, default=True Whether or not to shuffle the data before splitting. If shuffle=False then stratify must be None. stratify : array-like, default=None If not None, data is split in a stratified fashion, using this as the class labels. Read more in the :ref:`User Guide <stratification>`. Returns ------- splitting : list, length=2 * len(arrays) List containing train-test split of inputs. .. versionadded:: 0.16 If the input is sparse, the output will be a ``scipy.sparse.csr_matrix``. Else, output type is the same as the input type. Examples -------- >>> import numpy as np >>> from sklearn.model_selection import train_test_split >>> X, y = np.arange(10).reshape((5, 2)), range(5) >>> X array([[0, 1], [2, 3], [4, 5], [6, 7], [8, 9]]) >>> list(y) [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 array([[4, 5], [0, 1], [6, 7]]) >>> y_train [2, 0, 3] >>> X_test array([[2, 3], [8, 9]]) >>> y_test [1, 4] >>> train_test_split(y, shuffle=False) [[0, 1, 2], [3, 4]] """ n_arrays = len(arrays) if n_arrays == 0: raise ValueError("At least one array required as input") arrays = indexable(*arrays) n_samples = _num_samples(arrays[0]) n_train, n_test = _validate_shuffle_split( n_samples, test_size, train_size, default_test_size=0.25 ) if shuffle is False: if stratify is not None: raise ValueError( "Stratified train/test split is not implemented for shuffle=False" ) train = np.arange(n_train) test = np.arange(n_train, n_train + n_test) else: if stratify is not None: CVClass = StratifiedShuffleSplit else: CVClass = ShuffleSplit cv = CVClass(test_size=n_test, train_size=n_train, random_state=random_state) train, test = next(cv.split(X=arrays[0], y=stratify)) return list( chain.from_iterable( (_safe_indexing(a, train), _safe_indexing(a, test)) for a in arrays ) ) # Tell nose that train_test_split is not a test. # (Needed for external libraries that may use nose.) # Use setattr to avoid mypy errors when monkeypatching. setattr(train_test_split, "__test__", False) def _build_repr(self): # XXX This is copied from BaseEstimator's get_params cls = self.__class__ init = getattr(cls.__init__, "deprecated_original", cls.__init__) # Ignore varargs, kw and default values and pop self init_signature = signature(init) # Consider the constructor parameters excluding 'self' if init is object.__init__: args = [] else: args = sorted( [ p.name for p in init_signature.parameters.values() if p.name != "self" and p.kind != p.VAR_KEYWORD ] ) class_name = self.__class__.__name__ params = dict() for key in args: # We need deprecation warnings to always be on in order to # catch deprecated param values. # This is set in utils/__init__.py but it gets overwritten # when running under python3 somehow. warnings.simplefilter("always", FutureWarning) try: with warnings.catch_warnings(record=True) as w: value = getattr(self, key, None) if value is None and hasattr(self, "cvargs"): value = self.cvargs.get(key, None) if len(w) and w[0].category == FutureWarning: # if the parameter is deprecated, don't show it continue finally: warnings.filters.pop(0) params[key] = value return "%s(%s)" % (class_name, _pprint(params, offset=len(class_name))) def _yields_constant_splits(cv): # Return True if calling cv.split() always returns the same splits # We assume that if a cv doesn't have a shuffle parameter, it shuffles by # default (e.g. ShuffleSplit). If it actually doesn't shuffle (e.g. # LeaveOneOut), then it won't have a random_state parameter anyway, in # which case it will default to 0, leading to output=True shuffle = getattr(cv, "shuffle", True) random_state = getattr(cv, "random_state", 0) return isinstance(random_state, numbers.Integral) or not shuffle