Question

我想将数据拆分为训练，测试和验证数据集，这些数据集是分层的，但sklearn只提供cross_validation.train_test_split，它只能分为2个部分。如果我想这样做，我该怎么办

Answer 1

如果您想使用分层训练/测试分割，您可以使用StratifiedKFold in Sklearn

假设X是您的功能，y是您的标签，基于示例here：

from sklearn.model_selection import StratifiedKFold
cv_stf = StratifiedKFold(n_splits=3)
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]

更新：要将数据分成3个不同的百分比，使用numpy.split()可以这样做：

X_train, X_test, X_validate  = np.split(X, [int(.7*len(X)), int(.8*len(X))])
y_train, y_test, y_validate  = np.split(y, [int(.7*len(y)), int(.8*len(y))])

Answer 2

您也可以多次使用train_test_split来实现此目的。第二次，在第一次调用train_test_split的训练输出上运行它。

from sklearn.model_selection import train_test_split

def train_test_validate_stratified_split(features, targets, test_size=0.2, validate_size=0.1):
    # Get test sets
    features_train, features_test, targets_train, targets_test = train_test_split(
        features,
        targets,
        stratify=targets,
        test_size=test_size
    )
    # Run train_test_split again to get train and validate sets
    post_split_validate_size = validate_size / (1 - test_size)
    features_train, features_validate, targets_train, targets_validate = train_test_split(
        features_train,
        targets_train,
        stratify=targets_train,
        test_size=post_split_validate_size
    )
    return features_train, features_test, features_validate, targets_train, targets_test, targets_validate

如何使用sklearn

2 个答案: