xgboost - difference between XGBClassifier.feature_importances_ and XGBClassifier.get_booster().get_fscore()

Question

What is the difference between get_fscore() and feature_importances? Both are explained as feature importance but the importance values are different.

# model_smote = XGBClassifier()
# model_smote.fit(X_train_smote, y_train_smote)
model_smote = GridSearchCV(
    estimator = XGBClassifier(),
    param_grid=parameters,
    scoring='roc_auc',
    cv=5,
    verbose=2,
    return_train_score=True
)
model_smote.fit(X_train_smote, y_train_smote)
xgb = model_smote.best_estimator_

feature_importances = pd.Series(xgb.get_booster().get_fscore()).sort_values(ascending=False)
feature_importances.plot(kind='bar', title='Feature Importances', figsize=(10, 10))

get_fscore()

class xgboost.Booster(params=None, cache=(), model_file=None) - get_fscore(fmap='')

Get feature importance of each feature.

feature_importances_

class xgboost.XGBClassifier - property feature_importances_: numpy.ndarray

feature_importances = pd.DataFrame(
    data=xgb.feature_importances_,
    index=xgb.get_booster().feature_names,
    columns=['Importance']
).sort_values('Importance', ascending=False)
feature_importances.plot(kind='bar', title='Feature Importances', figsize=(10, 10))