FitFailedWarning：估计器拟合失败。这些参数在这个训练测试分区上的分数将设置为 nan答案

【问题标题】：FitFailedWarning: Estimator fit failed. The score on this train-test partition for these parameters will be set to nanFitFailedWarning：估计器拟合失败。这些参数在这个训练测试分区上的分数将设置为 nan
【发布时间】：2021-10-21 21:45:39
【问题描述】：

我正在尝试优化 XGB 回归模型的参数学习率和 max_depth：

from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import cross_val_score
from xgboost import XGBRegressor

param_grid = [
    # trying learning rates from 0.01 to 0.2
    {'eta ':[0.01, 0.05, 0.1, 0.2]},
    # and max depth from 4 to 10
    {'max_depth': [4, 6, 8, 10]}
  ]

xgb_model = XGBRegressor(random_state = 0)
grid_search = GridSearchCV(xgb_model, param_grid, cv=5,
                           scoring='neg_root_mean_squared_error',
                           return_train_score=True)

grid_search.fit(final_OH_X_train_scaled, y_train)

final_OH_X_train_scaled 是只包含数字特征的训练数据集。

y_train 是训练标签 - 也是数字。

这是返回错误：

FitFailedWarning: Estimator fit failed. The score on this train-test partition for these parameters will be set to nan.

我看过其他类似的问题，但还没有找到答案。

也尝试过：

param_grid = [
    # trying learning rates from 0.01 to 0.2
    # and max depth from 4 to 10
    {'eta ': [0.01, 0.05, 0.1, 0.2], 'max_depth': [4, 6, 8, 10]}   
  ]

但它会产生同样的错误。

编辑： 以下是数据示例：

final_OH_X_train_scaled.head()

y_train.head()

EDIT2：

可以通过以下方式检索数据样本：

final_OH_X_train_scaled = pd.DataFrame([[0.540617 ,1.204666 ,1.670791 ,-0.445424 ,-0.890944 ,-0.491098 ,0.094999 ,1.522411 ,-0.247443 ,-0.559572 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,1.0 ,0.0 ,0.0], 
                   [0.117467 ,-2.351903 ,0.718969 ,-0.119721 ,-0.874705 ,-0.530832 ,-1.385230 ,2.126612 ,-0.947731 ,-0.156967 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,1.0 ,0.0 ,0.0 ,0.0 ,0.0], 
                   [0.901138 ,-0.208256 ,-0.019134 ,0.265250 ,-0.889128 ,-0.467753 ,0.169306 ,-0.973256 ,0.056164 ,-0.671978 , 0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,1.0 ,0.0 ,0.0],
                   [2.074639 ,0.100602 ,-1.645121 ,0.929598 ,0.811911 ,1.364560 ,0.337242 ,0.435187 ,-0.388075 ,1.279959 , 0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,1.0], 
                   [2.198099 ,-0.496254 ,-0.917933 ,-1.418407 ,-0.975889 ,1.044495 ,0.254181 ,1.335285 ,2.079415 ,2.071974 , 0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,1.0 ,0.0 ,0.0 ,0.0 ,0.0]],
                  columns=['cont0' ,'cont1' ,'cont2' ,'cont3' ,'cont4' ,'cont5' ,'cont6' ,'cont7' ,'cont8' ,'cont9' ,'31' ,'32' ,'33' ,'34' ,'35' ,'36' ,'37' ,'38' ,'39' ,'40'])

【问题讨论】：

对我来说没有什么明显的错误。您能否发布几行您的final_OH_X_train_scaled 和y_train 数据，以便我们重现和调试？您的数据可能有问题。
@TCArlen 非常感谢您的反馈。请在上面查看我的编辑
太好了，谢谢。但是，为了在我的机器上进行检查和重现/调试，我需要将训练数据行/标签作为代码/数据，这样我就可以自己运行它。您可以将其发布为数据而不是屏幕截图吗？
链接中的数据不是上面截图中显示的方式转换的数据，来自final_OH_X_train_scaled.head()。请将这些值放入此示例问题中的代码中：stackoverflow.com/questions/68732791/… 您是否看到数据帧是如何从代码构造的，因此它是另一个机器上可重现的示例？谢谢
好的，请看上面

标签： python scikit-learn xgboost

【解决方案1】：

我能够重现该问题，但代码不适合，因为您的 eta 参数中有一个额外的空间！而不是这个：

{'eta ':[0.01, 0.05, 0.1, 0.2]},...

改成这样：

{'eta':[0.01, 0.05, 0.1, 0.2]},...

很遗憾，错误消息并没有太大帮助。

【讨论】：

【解决方案2】：

例如，如果对于LogisticRegression 你把网格设置成……

grid_lr = {
'cls__class_weight': [None, 'balanced'],
'cls__C': [0, .001, .01, .1, 1]
}

你会得到一个类似的错误；原因是C 只能采用正浮点值。因此，只需仔细检查超参数的命名或值就足以解决此问题。

【讨论】：