def explain_rf_feature_importance(estimator,
vec=None,
top=_TOP,
target_names=None, # ignored
targets=None, # ignored
feature_names=None,
feature_re=None,
feature_filter=None,
):
"""
Return an explanation of a tree-based ensemble estimator.
See :func:`eli5.explain_weights` for description of
``top``, ``feature_names``, ``feature_re`` and ``feature_filter``
parameters.
``target_names`` and ``targets`` parameters are ignored.
``vec`` is a vectorizer instance used to transform
raw features to the input of the estimator (e.g. a fitted
CountVectorizer instance); you can pass it instead of ``feature_names``.
"""
coef = estimator.feature_importances_
trees = np.array(estimator.estimators_).ravel()
coef_std = np.std([tree.feature_importances_ for tree in trees], axis=0)
return get_feature_importance_explanation(estimator, vec, coef,
coef_std=coef_std,
feature_names=feature_names,
feature_filter=feature_filter,
feature_re=feature_re,
top=top,
description=DESCRIPTION_RANDOM_FOREST,
is_regression=isinstance(estimator, RegressorMixin),
)
def explain_permutation_importance(estimator,
vec=None,
top=_TOP,
target_names=None, # ignored
targets=None, # ignored
feature_names=None,
feature_re=None,
feature_filter=None,
):
"""
Return an explanation of PermutationImportance.
See :func:`eli5.explain_weights` for description of
``top``, ``feature_names``, ``feature_re`` and ``feature_filter``
parameters.
``target_names`` and ``targets`` parameters are ignored.
``vec`` is a vectorizer instance used to transform
raw features to the input of the estimator (e.g. a fitted
CountVectorizer instance); you can pass it instead of ``feature_names``.
"""
coef = estimator.feature_importances_
coef_std = estimator.feature_importances_std_
return get_feature_importance_explanation(estimator, vec, coef,
coef_std=coef_std,
feature_names=feature_names,
feature_filter=feature_filter,
feature_re=feature_re,
top=top,
description=DESCRIPTION_SCORE_DECREASE + estimator.caveats_,
is_regression=isinstance(estimator.wrapped_estimator_, RegressorMixin),
)
def explain_weights_xgboost(
xgb,
vec=None,
top=20,
target_names=None, # ignored
targets=None, # ignored
feature_names=None,
feature_re=None, # type: Pattern[str]
feature_filter=None,
importance_type='gain',
):
"""
Return an explanation of an XGBoost estimator (via scikit-learn wrapper
XGBClassifier or XGBRegressor, or via xgboost.Booster)
as feature importances.
See :func:`eli5.explain_weights` for description of
``top``, ``feature_names``,
``feature_re`` and ``feature_filter`` parameters.
``target_names`` and ``targets`` parameters are ignored.
Parameters
----------
importance_type : str, optional
A way to get feature importance. Possible values are:
- 'gain' - the average gain of the feature when it is used in trees
(default)
- 'weight' - the number of times a feature is used to split the data
across all trees
- 'cover' - the average coverage of the feature when it is used in trees
"""
booster, is_regression = _check_booster_args(xgb)
xgb_feature_names = booster.feature_names
coef = _xgb_feature_importances(booster, importance_type=importance_type)
return get_feature_importance_explanation(
xgb,
vec,
coef,
feature_names=feature_names,
estimator_feature_names=xgb_feature_names,
feature_filter=feature_filter,
feature_re=feature_re,
top=top,
description=DESCRIPTION_XGBOOST,
is_regression=is_regression,
num_features=coef.shape[-1],
)
def explain_weights_catboost(catb,
vec=None,
top=20,
importance_type='PredictionValuesChange',
feature_names=None,
pool=None):
"""
Return an explanation of an CatBoost estimator (CatBoostClassifier,
CatBoost, CatBoostRegressor) as feature importances.
See :func:`eli5.explain_weights` for description of
``top``, ``feature_names``,
``feature_re`` and ``feature_filter`` parameters.
``target_names`` and ``targets`` parameters are ignored.
Parameters
----------
:param 'importance_type' : str, optional
A way to get feature importance. Possible values are:
- 'PredictionValuesChange' (default) - The individual importance
values for each of the input features.
- 'LossFunctionChange' - The individual importance values for
each of the input features for ranking metrics
(requires training data to be passed or a similar dataset with Pool)
:param 'pool' : catboost.Pool, optional
To be passed if explain_weights_catboost has importance_type set
to LossFunctionChange. The catboost feature_importances uses the Pool
datatype to calculate the parameter for the specific importance_type.
"""
is_regression = _is_regression(catb)
catb_feature_names = catb.feature_names_
coef = _catb_feature_importance(catb,
importance_type=importance_type,
pool=pool)
return get_feature_importance_explanation(
catb,
vec,
coef,
feature_names=feature_names,
estimator_feature_names=catb_feature_names,
feature_filter=None,
feature_re=None,
top=top,
description=DESCRIPTION_CATBOOST,
is_regression=is_regression,
num_features=coef.shape[-1])
def explain_weights_lightgbm(
lgb,
vec=None,
top=20,
target_names=None, # ignored
targets=None, # ignored
feature_names=None,
feature_re=None,
feature_filter=None,
importance_type='gain',
):
"""
Return an explanation of an LightGBM estimator (via scikit-learn wrapper
LGBMClassifier or LGBMRegressor) as feature importances.
See :func:`eli5.explain_weights` for description of
``top``, ``feature_names``,
``feature_re`` and ``feature_filter`` parameters.
``target_names`` and ``targets`` parameters are ignored.
Parameters
----------
importance_type : str, optional
A way to get feature importance. Possible values are:
- 'gain' - the average gain of the feature when it is used in trees
(default)
- 'split' - the number of times a feature is used to split the data
across all trees
- 'weight' - the same as 'split', for compatibility with xgboost
"""
coef = _get_lgb_feature_importances(lgb, importance_type)
lgb_feature_names = lgb.booster_.feature_name()
return get_feature_importance_explanation(
lgb,
vec,
coef,
feature_names=feature_names,
estimator_feature_names=lgb_feature_names,
feature_filter=feature_filter,
feature_re=feature_re,
top=top,
description=DESCRIPTION_LIGHTGBM,
num_features=coef.shape[-1],
is_regression=isinstance(lgb, lightgbm.LGBMRegressor),
)
