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

Question

I'm trying to optimize the parameters learning rate and max_depth of a XGB regression model:

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 is the training dataset that contains only numerical features.

y_train is the training label - also numerical.

This is returning the error:

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

I've seen other similar questions, but couldn't find an answer yet.

Also tried with:

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]}   
  ]

But it generates the same error.

EDIT: Here's a sample of the data:

final_OH_X_train_scaled.head()

y_train.head()

EDIT2:

The data sample may be retrieved with:

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'])

Answer 1

I was able to reproduce the problem and the code fails to fit because there is an extra space in your eta parameter! Instead of this:

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

Change it to this:

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

The error message was unfortunately not very helpful.

Answer 2

Also for example, if for a LogisticRegression you set the grid to sth like

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

You'll get a similar error; the reason being that C could only take positive float values. Hence, simply double checking the naming or the values of the hyperparameters should be enough to resolve this issue.