使用GroupShuffleSplit做嵌套交叉验证遇Pickle生成器错误求助

问题背景 #

使用GroupShuffleSplit实现分组拆分的嵌套交叉验证时，将group_split.split(...)返回的生成器传入GridSearchCV和cross_val_score，触发TypeError: cannot pickle 'generator' object错误。核心原因是生成器无法被序列化，而scikit-learn并行计算过程中需要对CV对象进行序列化操作。

错误代码示例 #

import numpy as np
from sklearn.model_selection import GroupShuffleSplit, GridSearchCV, cross_val_score
from sklearn.ensemble import RandomForestClassifier

X = np.random.rand(100, 10)
y = np.random.randint(2, size=100)
groups = np.random.randint(4, size=100)  # 示例分组标签

rf_classifier = RandomForestClassifier()
param_grid = {'n_estimators': [50, 100, 200]}

inner_cv = GroupShuffleSplit(n_splits=5, test_size=0.2)
outer_cv = GroupShuffleSplit(n_splits=5, test_size=0.2)

# 错误写法：传入split()返回的生成器
grid_search = GridSearchCV(estimator=rf_classifier, param_grid=param_grid, cv=inner_cv.split(X, y, groups=groups))
nested_scores = cross_val_score(estimator=grid_search, X=X, y=y, cv=outer_cv.split(X, y, groups=groups))

解决方案 #

无需手动调用split()生成拆分器，直接将GroupShuffleSplit实例传入cv参数即可。scikit-learn会自动在内部处理split()调用和分组逻辑，同时保证对象可序列化。

修正后的代码：

import numpy as np
from sklearn.model_selection import GroupShuffleSplit, GridSearchCV, cross_val_score
from sklearn.ensemble import RandomForestClassifier

X = np.random.rand(100, 10)
y = np.random.randint(2, size=100)
groups = np.random.randint(4, size=100)  # 示例分组标签

rf_classifier = RandomForestClassifier()
param_grid = {'n_estimators': [50, 100, 200]}

# 直接传入GroupShuffleSplit实例，而非split()生成器
inner_cv = GroupShuffleSplit(n_splits=5, test_size=0.2)
outer_cv = GroupShuffleSplit(n_splits=5, test_size=0.2)

grid_search = GridSearchCV(estimator=rf_classifier, param_grid=param_grid, cv=inner_cv)
# cross_val_score自动传递groups参数给outer_cv的split方法
nested_scores = cross_val_score(estimator=grid_search, X=X, y=y, groups=groups, cv=outer_cv)

print(nested_scores)

关键说明 #

• GroupShuffleSplit类本身是可序列化的，而split()返回的生成器不支持序列化，这是错误的核心诱因。
• 调用cross_val_score时传入groups参数，函数会自动将其传递给外层CV对象的split()方法，实现分组拆分逻辑。
• 内层GridSearchCV运行时，会自动从外层交叉验证的训练集中获取对应分组子集，无需额外配置。

内容的提问来源于stack exchange，提问作者Sebastian_Ulrich