为什么我得到了一个 TypeError: __init__() 在 python sklearn 中得到了一个意外的关键字参数“outer_cv”？

Question

我已经运行了链接中的示例代码： https : //github.com/casperbh96/Nested-Cross-Validation/blob/master/Example%20Notebook%20-%20NestedCV.ipynb ，但出现了一个错误： init ()得到一个意外的关键字参数'outer_cv'，我查看了源代码，'outer_cv'包含​​在int ()中，如何解决？ 示例代码也粘贴如下：

from nested_cv import NestedCV

import pandas as pd
import numpy as np
from sklearn.datasets import load_boston, load_iris, load_breast_cancer
from sklearn.ensemble import RandomForestRegressor, RandomForestClassifier
from sklearn.model_selection import KFold

# When using Random Search, we get a user warning with this little number of hyperparameters
# Suppress it
import warnings
warnings.simplefilter(action='ignore', category=UserWarning)
warnings.simplefilter(action='ignore', category=FutureWarning)

boston = load_boston()
X = boston.data
y = boston.target

# Define a parameters grid
param_grid = {
     'max_depth': [3, 7, 10, None],
     'n_estimators': [100,200],
     'min_samples_split':[2,3,5,7,10]
}

# Either specify a strategy or number
# Here we choose a strategy
outer_cv = KFold(n_splits=5,
                 shuffle=True,
                 random_state=123)
inner_cv = KFold(n_splits=5,
                 shuffle=True,
                 random_state=123)

NCV = NestedCV(model=RandomForestRegressor(), params_grid=param_grid,
               outer_cv=outer_cv, inner_cv=inner_cv, n_jobs = -1,
               cv_options={'sqrt_of_score':True, 
                           'recursive_feature_elimination':False, 
                           'rfe_n_features':2})
NCV.fit(X=X,y=y)

NCV.outer_scores

Answer 1

这应该让你更接近，虽然我不能让 NCV.fit 在任何合理的时间内完成：

from nested_cv import NestedCV

import pandas as pd
import numpy as np
from sklearn.datasets import load_boston, load_iris, load_breast_cancer
from sklearn.ensemble import RandomForestRegressor, RandomForestClassifier
from sklearn.model_selection import KFold


# When using Random Search, we get a user warning with this little number of hyperparameters
# Suppress it
import warnings
warnings.simplefilter(action='ignore', category=UserWarning)
warnings.simplefilter(action='ignore', category=FutureWarning)

boston = load_boston()
X = boston.data
y = boston.target

# Define a parameters grid
param_grid = {
     'max_depth': [3, 7, 10, None],
     'n_estimators': [100,200],
     'min_samples_split':[2,3,5,7,10]
}

# Either specify a strategy or number
# Here we choose a strategy
outer_cv = KFold(n_splits=5,
                 shuffle=True,
                 random_state=123)
inner_cv = KFold(n_splits=5,
                 shuffle=True,
                 random_state=123)


NCV = NestedCV(model=RandomForestRegressor(), params_grid=param_grid,
               outer_kfolds=outer_cv.n_splits, inner_kfolds=inner_cv.n_splits, n_jobs = -1,
               cv_options={'sqrt_of_score':True, 
                           'recursive_feature_elimination':False, 
                           'rfe_n_features':2})
#NCV.fit(X=X,y=y)

NCV.outer_scores