XGBClassifier fails when I pass early_stopping_rounds

Question

I am using xgboost in the following way:

from xgboost import XGBClassifier
clf = XGBClassifier()
clf = clf.fit(df_train, df_train_labels, verbose=True)

This works well. However, if I add an early_stopping_rounds parameter, like this:

clf = clf.fit(df_train, df_train_labels, early_stopping_rounds=10, verbose=True)

I get this error:

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-16-786925228ae5> in <module>()
      9 
     10 
---> 11 clf = clf.fit(df_train, df_train_labels, early_stopping_rounds=10, verbose=True)
     12 print("after fit")
     13 prediction = np.exp(clf.predict(df_test))

~/anaconda3/envs/python3/lib/python3.6/site-packages/xgboost/sklearn.py in fit(self, X, y, sample_weight, eval_set, eval_metric, early_stopping_rounds, verbose)
    443                               early_stopping_rounds=early_stopping_rounds,
    444                               evals_result=evals_result, obj=obj, feval=feval,
--> 445                               verbose_eval=verbose)
    446 
    447         self.objective = xgb_options["objective"]

~/anaconda3/envs/python3/lib/python3.6/site-packages/xgboost/training.py in train(params, dtrain, num_boost_round, evals, obj, feval, maximize, early_stopping_rounds, evals_result, verbose_eval, learning_rates, xgb_model, callbacks)
    203                            evals=evals,
    204                            obj=obj, feval=feval,
--> 205                            xgb_model=xgb_model, callbacks=callbacks)
    206 
    207 

~/anaconda3/envs/python3/lib/python3.6/site-packages/xgboost/training.py in _train_internal(params, dtrain, num_boost_round, evals, obj, feval, xgb_model, callbacks)
     99                                end_iteration=num_boost_round,
    100                                rank=rank,
--> 101                                evaluation_result_list=evaluation_result_list))
    102         except EarlyStopException:
    103             break

~/anaconda3/envs/python3/lib/python3.6/site-packages/xgboost/callback.py in callback(env)
    190     def callback(env):
    191         """internal function"""
--> 192         score = env.evaluation_result_list[-1][1]
    193         if len(state) == 0:
    194             init(env)

IndexError: list index out of range

I looked this up and I saw that the fit method can be passed a multitude of parameters, so I don't believe that the fact that I added early_stopping_rounds should cause problems.

Any idea what could be the cause of this error?

Answer 1

The reason for this error is that you haven't specified an eval_set, which xgboost uses to determine when to stop for early-stopping.

See the docs for the fit method here.

eval_set (list, optional) – A list of (X, y) tuple pairs to use as a validation set for early-stopping

For example, if you've split your data into train and test sets, you would use something along these lines:

eval_set = [(X_test, y_test)]

clf = clf.fit(df_train,
              df_train_labels,
              eval_set=eval_set,
              early_stopping_rounds=10,
              verbose=True)