def test_importance(self, iris_X, iris_y, importance, scale_permutation_importance, local_importance):
rfc = RangerForestClassifier(
importance=importance,
scale_permutation_importance=scale_permutation_importance,
local_importance=local_importance,
)
if importance not in ["none", "impurity", "impurity_corrected", "permutation"]:
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
return
rfc.fit(iris_X, iris_y)
if importance == "none":
assert rfc.importance_mode_ == 0
elif importance == "impurity":
assert rfc.importance_mode_ == 1
elif importance == "impurity_corrected":
assert rfc.importance_mode_ == 5
elif importance == "permutation":
if local_importance:
assert rfc.importance_mode_ == 6
elif scale_permutation_importance:
assert rfc.importance_mode_ == 2
else:
assert rfc.importance_mode_ == 3
def test_split_rule(self, iris_X, iris_y, split_rule):
rfc = RangerForestClassifier(split_rule=split_rule)
if split_rule not in ["gini", "extratrees", "hellinger"]:
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
return
# hellinger can only be used in binary classification
if split_rule == "hellinger":
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
iris_2 = [0 if v == 2 else v for v in iris_y]
rfc.fit(iris_X, iris_2)
if split_rule == "gini":
assert rfc.split_rule_ == 1
elif split_rule == "extratrees":
assert rfc.split_rule_ == 5
if split_rule == "hellinger":
assert rfc.split_rule_ == 7
if split_rule == "extratrees":
rfc = RangerForestClassifier(
split_rule=split_rule, respect_categorical_features="partition", save_memory=True
)
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
else:
rfc = RangerForestClassifier(split_rule=split_rule, num_random_splits=2)
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
def test_categorical_features(self, iris_X, iris_y,
respect_categorical_features):
# add a categorical feature
categorical_col = np.atleast_2d(
np.array([random.choice([0, 1]) for _ in range(iris_X.shape[0])]))
iris_X_c = np.hstack((iris_X, categorical_col.transpose()))
categorical_features = [iris_X.shape[1]]
rfc = RangerForestClassifier(
respect_categorical_features=respect_categorical_features,
categorical_features=categorical_features,
)
if respect_categorical_features not in [
"partition", "ignore", "order"
]:
with pytest.raises(ValueError):
rfc.fit(iris_X_c, iris_y)
return
rfc.fit(iris_X_c, iris_y)
if respect_categorical_features in ("ignore", "order"):
assert rfc.categorical_features_ == []
else:
assert rfc.categorical_features_ == [
str(c).encode() for c in categorical_features
]
def test_verbose(self, iris_X, iris_y, verbose, capfd):
rfc = RangerForestClassifier(verbose=verbose)
rfc.fit(iris_X, iris_y)
captured = capfd.readouterr()
if verbose:
assert len(captured.out) > 0
else:
assert len(captured.out) == 0
def test_sample_fraction_replace(self, iris_X, iris_y, replace):
rfc = RangerForestClassifier(replace=replace)
rfc.fit(iris_X, iris_y)
if replace:
assert rfc.sample_fraction_ == [1.0]
else:
assert rfc.sample_fraction_ == [0.632]
def test_serialize(self, iris_X, iris_y):
tf = tempfile.TemporaryFile()
rfc = RangerForestClassifier()
rfc.fit(iris_X, iris_y)
pickle.dump(rfc, tf)
tf.seek(0)
new_rfc = pickle.load(tf)
pred = new_rfc.predict(iris_X)
assert len(pred) == iris_X.shape[0]
def test_predict_log_proba(self, iris_X, iris_y):
rfc = RangerForestClassifier()
rfc.fit(iris_X, iris_y)
pred = rfc.predict_log_proba(iris_X)
assert len(pred) == iris_X.shape[0]
# test with single record
iris_X_record = iris_X[0:1, :]
pred = rfc.predict_log_proba(iris_X_record)
assert len(pred) == 1
def test_shap_classifier(iris_X, iris_y):
from shap import TreeExplainer
forest = RangerForestClassifier(enable_tree_details=True)
forest.fit(iris_X, iris_y)
with shap_patch():
explainer = TreeExplainer(model=forest)
shap_values = explainer.shap_values(iris_X)
print(shap_values)
def test_predict(self, iris_X, iris_y):
forest = RangerForestClassifier()
forest.fit(iris_X, iris_y)
pred = forest.predict(iris_X)
assert len(pred) == iris_X.shape[0]
# test with single record
iris_X_record = iris_X[0:1, :]
pred = forest.predict(iris_X_record)
assert len(pred) == 1
def test_fit(self, iris_X, iris_y):
rfc = RangerForestClassifier()
with pytest.raises(NotFittedError):
check_is_fitted(rfc)
rfc.fit(iris_X, iris_y)
check_is_fitted(rfc)
assert hasattr(rfc, "classes_")
assert hasattr(rfc, "n_classes_")
assert hasattr(rfc, "ranger_forest_")
assert hasattr(rfc, "ranger_class_order_")
assert hasattr(rfc, "n_features_in_")
def test_regularization(self, iris_X, iris_y):
rfc = RangerForestClassifier()
rfc.fit(iris_X, iris_y)
assert rfc.regularization_factor_ == []
assert not rfc.use_regularization_factor_
# vector must be between 0 and 1 and length matching feature num
for r in [[1.1], [-0.1], [1, 1]]:
rfc = RangerForestClassifier(regularization_factor=r)
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
# vector of ones isn't applied
rfc = RangerForestClassifier(regularization_factor=[1] * iris_X.shape[1])
rfc.fit(iris_X, iris_y)
assert rfc.regularization_factor_ == []
assert not rfc.use_regularization_factor_
# regularization vector is used
reg = [0.5]
rfc = RangerForestClassifier(regularization_factor=reg, n_jobs=2)
# warns if n_jobs is not one since parallelization can't be used
with pytest.warns(Warning):
rfc.fit(iris_X, iris_y)
assert rfc.n_jobs_ == 1
assert rfc.regularization_factor_ == reg
assert rfc.use_regularization_factor_
def test_importance_pvalues(self, iris_X_mod, iris_y, importance, mod):
rfc = RangerForestClassifier(importance=importance)
np.random.seed(42)
if importance not in [
"none", "impurity", "impurity_corrected", "permutation"
]:
with pytest.raises(ValueError):
rfc.fit(iris_X_mod, iris_y)
return
if not importance == "impurity_corrected":
rfc.fit(iris_X_mod, iris_y)
with pytest.raises(ValueError):
rfc.get_importance_pvalues()
return
# Test error for no non-negative importance values
if mod == "none":
rfc.fit(iris_X_mod, iris_y)
with pytest.raises(ValueError):
rfc.get_importance_pvalues()
return
rfc.fit(iris_X_mod, iris_y)
assert len(rfc.get_importance_pvalues()) == iris_X_mod.shape[1]
def test_estimators_(self, iris_X, iris_y):
forest = RangerForestClassifier(n_estimators=10)
with pytest.raises(AttributeError):
_ = forest.estimators_
forest.fit(iris_X, iris_y)
with pytest.raises(ValueError):
_ = forest.estimators_
forest = RangerForestClassifier(n_estimators=10,
enable_tree_details=True)
forest.fit(iris_X, iris_y)
estimators = forest.estimators_
assert len(estimators) == 10
assert isinstance(estimators[0], RangerTreeClassifier)
check_is_fitted(estimators[0])
def test_get_estimator(self, iris_X, iris_y):
forest = RangerForestClassifier(n_estimators=10)
with pytest.raises(NotFittedError):
_ = forest.get_estimator(idx=0)
forest.fit(iris_X, iris_y)
with pytest.raises(ValueError):
_ = forest.get_estimator(0)
forest = RangerForestClassifier(n_estimators=10,
enable_tree_details=True)
forest.fit(iris_X, iris_y)
estimator = forest.get_estimator(0)
estimator.predict(iris_X)
assert isinstance(estimator, RangerTreeClassifier)
with pytest.raises(IndexError):
_ = forest.get_estimator(idx=20)
def test_sample_fraction(self, iris_X, iris_y):
rfc = RangerForestClassifier(sample_fraction=[0.69])
rfc.fit(iris_X, iris_y)
assert rfc.sample_fraction_ == [0.69]
rfc = RangerForestClassifier(sample_fraction=0.69)
rfc.fit(iris_X, iris_y)
assert rfc.sample_fraction_ == [0.69]
# test with single record
iris_X_record = iris_X[0:1, :]
pred = rfc.predict(iris_X_record)
assert len(pred) == 1
pred = rfc.predict_proba(iris_X_record)
assert len(pred) == 1
pred = rfc.predict_log_proba(iris_X_record)
assert len(pred) == 1
def test_feature_importances_(self, iris_X, iris_y, importance, local_importance):
rfc = RangerForestClassifier(importance=importance, local_importance=local_importance)
with pytest.raises(AttributeError):
_ = rfc.feature_importances_
if importance == "INVALID":
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
return
rfc.fit(iris_X, iris_y)
if importance == "none":
with pytest.raises(ValueError):
_ = rfc.feature_importances_
else:
assert len(rfc.feature_importances_) == iris_X.shape[1]
def test_accuracy(self, iris_X, iris_y):
X_train, X_test, y_train, y_test = train_test_split(iris_X, iris_y, test_size=0.33, random_state=42)
# train and test a random forest classifier
rf = RandomForestClassifier()
rf.fit(X_train, y_train)
y_pred_rf = rf.predict(X_test)
rf_acc = accuracy_score(y_test, y_pred_rf)
# train and test a ranger classifier
ra = RangerForestClassifier()
ra.fit(X_train, y_train)
y_pred_ra = ra.predict(X_test)
ranger_acc = accuracy_score(y_test, y_pred_ra)
# the accuracy should be good
assert rf_acc > 0.9
assert ranger_acc > 0.9
def test_class_weights(self, iris_X, iris_y):
X_train, X_test, y_train, y_test = train_test_split(iris_X,
iris_y,
test_size=0.5,
random_state=42)
forest = RangerForestClassifier()
weights = {
0: 0.7,
1: 0.2,
2: 0.1,
}
forest.fit(X_train, y_train, class_weights=weights)
forest.predict(X_test)
forest = RangerForestClassifier()
m = {0: "a", 1: "b", 2: "c"}
y_train_str = [m.get(v) for v in y_train]
weights = {
"a": 0.7,
"b": 0.2,
"c": 0.1,
}
forest.fit(X_train, y_train_str, class_weights=weights)
forest.predict(X_test)
weights = {
0: 0.7,
}
with pytest.raises(ValueError):
forest.fit(X_train, y_train, class_weights=weights)
def test_categorical_features(self, iris_X, iris_y,
respect_categorical_features):
# add a categorical feature
categorical_col = np.atleast_2d(
np.array([random.choice([0, 1]) for _ in range(iris_X.shape[0])]))
iris_X_c = np.hstack((iris_X, categorical_col.transpose()))
categorical_features = [iris_X.shape[1]]
forest = RangerForestClassifier(
respect_categorical_features=respect_categorical_features, )
if respect_categorical_features not in [
"partition", "ignore", "order"
]:
with pytest.raises(ValueError):
forest.fit(iris_X_c,
iris_y,
categorical_features=categorical_features)
return
forest.fit(iris_X_c, iris_y, categorical_features=categorical_features)
forest.predict(iris_X_c)
def test_inbag(self, iris_X, iris_y):
inbag = [[1, 2, 3], [2, 3, 4]]
rfc = RangerForestClassifier(n_estimators=2, inbag=inbag)
rfc.fit(iris_X, iris_y)
# inbag list different length from n_estimators
rfc = RangerForestClassifier(n_estimators=1, inbag=inbag)
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
# can't use inbag with sample weight
rfc = RangerForestClassifier(inbag=inbag)
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y, sample_weight=[1] * len(iris_y))
# can't use class sampling and inbag
rfc = RangerForestClassifier(inbag=inbag, sample_fraction=[1, 1])
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
def test_split_select_weights(self, iris_X, iris_y):
n_trees = 10
weights = [0.1] * iris_X.shape[1]
forest = RangerForestClassifier(n_estimators=n_trees)
forest.fit(iris_X, iris_y, split_select_weights=weights)
weights = [0.1] * (iris_X.shape[1] - 1)
forest = RangerForestClassifier(n_estimators=n_trees)
with pytest.raises(RuntimeError):
forest.fit(iris_X, iris_y, split_select_weights=weights)
weights = [[0.1] * (iris_X.shape[1])] * n_trees
forest = RangerForestClassifier(n_estimators=n_trees)
forest.fit(iris_X, iris_y, split_select_weights=weights)
weights = [[0.1] * (iris_X.shape[1])] * (n_trees + 1)
forest = RangerForestClassifier(n_estimators=n_trees)
with pytest.raises(RuntimeError):
forest.fit(iris_X, iris_y, split_select_weights=weights)
def test_mtry(self, iris_X, iris_y, mtry):
rfc = RangerForestClassifier(mtry=mtry)
if callable(mtry) and mtry(5) > 5:
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
return
elif not callable(mtry) and (mtry < 0 or mtry > iris_X.shape[0]):
with pytest.raises(ValueError):
rfc.fit(iris_X, iris_y)
return
rfc.fit(iris_X, iris_y)
if callable(mtry):
assert rfc.mtry_ == mtry(iris_X.shape[1])
else:
assert rfc.mtry_ == mtry
def test_clone(self, iris_X, iris_y):
rfc = RangerForestClassifier()
rfc.fit(iris_X, iris_y)
clone(rfc)
def test_clone(self, iris_X, iris_y):
forest = RangerForestClassifier()
forest.fit(iris_X, iris_y)
clone(forest)
def RangerForestIMPL(train_embeddings, train_labels,test_embeddings,test_labels,single_label=True):
rfc = RangerForestClassifier()
rfc.fit(train_embeddings, train_labels)
predsranger = rfc.predict(test_embeddings)
if single_label==True:
return single_label_metrics(test_labels,predsranger)
def test_always_split_features(self, iris_X, iris_y):
rfc = RangerForestClassifier(always_split_features=[0])
rfc.fit(iris_X, iris_y)
# feature 0 is in every tree split
for tree in rfc.ranger_forest_["forest"]["split_var_ids"]:
assert 0 in tree