Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Commit 48e826f

Browse files
jmschreijmschrei
committed
FIX unit tests
1 parent 55203ed commit 48e826f

File tree

1 file changed

+91
-99
lines changed

1 file changed

+91
-99
lines changed

‎sklearn/ensemble/tests/test_forest.py

Lines changed: 91 additions & 99 deletions
Original file line numberDiff line numberDiff line change
@@ -90,21 +90,20 @@ def check_classification_toy(name):
9090
"""Check classification on a toy dataset."""
9191
ForestClassifier = FOREST_CLASSIFIERS[name]
9292

93-
for presort in True, False:
94-
clf = ForestClassifier(n_estimators=10, random_state=1, presort=presort)
95-
clf.fit(X, y)
96-
assert_array_equal(clf.predict(T), true_result)
97-
assert_equal(10, len(clf))
93+
clf = ForestClassifier(n_estimators=10, random_state=1)
94+
clf.fit(X, y)
95+
assert_array_equal(clf.predict(T), true_result)
96+
assert_equal(10, len(clf))
9897

99-
clf = ForestClassifier(n_estimators=10, max_features=1,
100-
random_state=1, presort=presort)
101-
clf.fit(X, y)
102-
assert_array_equal(clf.predict(T), true_result)
103-
assert_equal(10, len(clf))
98+
clf = ForestClassifier(n_estimators=10, max_features=1,
99+
random_state=1)
100+
clf.fit(X, y)
101+
assert_array_equal(clf.predict(T), true_result)
102+
assert_equal(10, len(clf))
104103

105-
# also test apply
106-
leaf_indices = clf.apply(X)
107-
assert_equal(leaf_indices.shape, (len(X), clf.n_estimators))
104+
# also test apply
105+
leaf_indices = clf.apply(X)
106+
assert_equal(leaf_indices.shape, (len(X), clf.n_estimators))
108107

109108

110109
def test_classification_toy():
@@ -116,20 +115,19 @@ def check_iris_criterion(name, criterion):
116115
# Check consistency on dataset iris.
117116
ForestClassifier = FOREST_CLASSIFIERS[name]
118117

119-
for presort in True, False:
120-
clf = ForestClassifier(n_estimators=10, criterion=criterion,
121-
random_state=1, presort=presort)
122-
clf.fit(iris.data, iris.target)
123-
score = clf.score(iris.data, iris.target)
124-
assert_greater(score, 0.9, "Failed with criterion %s and score = %f"
125-
% (criterion, score))
118+
clf = ForestClassifier(n_estimators=10, criterion=criterion,
119+
random_state=1)
120+
clf.fit(iris.data, iris.target)
121+
score = clf.score(iris.data, iris.target)
122+
assert_greater(score, 0.9, "Failed with criterion %s and score = %f"
123+
% (criterion, score))
126124

127-
clf = ForestClassifier(n_estimators=10, criterion=criterion,
128-
max_features=2, random_state=1, presort=presort)
129-
clf.fit(iris.data, iris.target)
130-
score = clf.score(iris.data, iris.target)
131-
assert_greater(score, 0.5, "Failed with criterion %s and score = %f"
132-
% (criterion, score))
125+
clf = ForestClassifier(n_estimators=10, criterion=criterion,
126+
max_features=2, random_state=1)
127+
clf.fit(iris.data, iris.target)
128+
score = clf.score(iris.data, iris.target)
129+
assert_greater(score, 0.5, "Failed with criterion %s and score = %f"
130+
% (criterion, score))
133131

134132

135133
def test_iris():
@@ -141,20 +139,19 @@ def check_boston_criterion(name, criterion):
141139
# Check consistency on dataset boston house prices.
142140
ForestRegressor = FOREST_REGRESSORS[name]
143141

144-
for presort in True, False:
145-
clf = ForestRegressor(n_estimators=5, criterion=criterion,
146-
random_state=1, presort=presort)
147-
clf.fit(boston.data, boston.target)
148-
score = clf.score(boston.data, boston.target)
149-
assert_greater(score, 0.95, "Failed with max_features=None, criterion %s "
150-
"and score = %f" % (criterion, score))
142+
clf = ForestRegressor(n_estimators=5, criterion=criterion,
143+
random_state=1)
144+
clf.fit(boston.data, boston.target)
145+
score = clf.score(boston.data, boston.target)
146+
assert_greater(score, 0.95, "Failed with max_features=None, criterion %s "
147+
"and score = %f" % (criterion, score))
151148

152-
clf = ForestRegressor(n_estimators=5, criterion=criterion,
153-
max_features=6, random_state=1, presort=presort)
154-
clf.fit(boston.data, boston.target)
155-
score = clf.score(boston.data, boston.target)
156-
assert_greater(score, 0.95, "Failed with max_features=6, criterion %s "
157-
"and score = %f" % (criterion, score))
149+
clf = ForestRegressor(n_estimators=5, criterion=criterion,
150+
max_features=6, random_state=1)
151+
clf.fit(boston.data, boston.target)
152+
score = clf.score(boston.data, boston.target)
153+
assert_greater(score, 0.95, "Failed with max_features=6, criterion %s "
154+
"and score = %f" % (criterion, score))
158155

159156

160157
def test_boston():
@@ -199,38 +196,37 @@ def test_probability():
199196
def check_importances(X, y, name, criterion):
200197
ForestEstimator = FOREST_ESTIMATORS[name]
201198

202-
for presort in True, False:
203-
est = ForestEstimator(n_estimators=20, criterion=criterion,
204-
random_state=0, presort=presort)
205-
est.fit(X, y)
206-
importances = est.feature_importances_
207-
n_important = np.sum(importances > 0.1)
208-
assert_equal(importances.shape[0], 10)
209-
assert_equal(n_important, 3)
210-
211-
X_new = est.transform(X, threshold="mean")
212-
assert_less(X_new.shape[1], X.shape[1])
213-
214-
# Check with parallel
215-
importances = est.feature_importances_
216-
est.set_params(n_jobs=2)
217-
importances_parrallel = est.feature_importances_
218-
assert_array_almost_equal(importances, importances_parrallel)
219-
220-
# Check with sample weights
221-
sample_weight = check_random_state(0).randint(1, 10, len(X))
199+
est = ForestEstimator(n_estimators=20, criterion=criterion,
200+
random_state=0)
201+
est.fit(X, y)
202+
importances = est.feature_importances_
203+
n_important = np.sum(importances > 0.1)
204+
assert_equal(importances.shape[0], 10)
205+
assert_equal(n_important, 3)
206+
207+
X_new = est.transform(X, threshold="mean")
208+
assert_less(X_new.shape[1], X.shape[1])
209+
210+
# Check with parallel
211+
importances = est.feature_importances_
212+
est.set_params(n_jobs=2)
213+
importances_parrallel = est.feature_importances_
214+
assert_array_almost_equal(importances, importances_parrallel)
215+
216+
# Check with sample weights
217+
sample_weight = check_random_state(0).randint(1, 10, len(X))
218+
est = ForestEstimator(n_estimators=20, random_state=0,
219+
criterion=criterion)
220+
est.fit(X, y, sample_weight=sample_weight)
221+
importances = est.feature_importances_
222+
assert_true(np.all(importances >= 0.0))
223+
224+
for scale in [0.5, 10, 100]:
222225
est = ForestEstimator(n_estimators=20, random_state=0,
223-
criterion=criterion, presort=presort)
224-
est.fit(X, y, sample_weight=sample_weight)
225-
importances = est.feature_importances_
226-
assert_true(np.all(importances >= 0.0))
227-
228-
for scale in [0.5, 10, 100]:
229-
est = ForestEstimator(n_estimators=20, random_state=0,
230-
criterion=criterion, presort=presort)
231-
est.fit(X, y, sample_weight=scale * sample_weight)
232-
importances_bis = est.feature_importances_
233-
assert_less(np.abs(importances - importances_bis).mean(), 0.001)
226+
criterion=criterion)
227+
est.fit(X, y, sample_weight=scale * sample_weight)
228+
importances_bis = est.feature_importances_
229+
assert_less(np.abs(importances - importances_bis).mean(), 0.001)
234230

235231

236232
def test_importances():
@@ -325,20 +321,18 @@ def mdi_importance(X_m, X, y):
325321
for i in range(n_features):
326322
true_importances[i] = mdi_importance(i, X, y)
327323

328-
for presort in True, False:
329-
# Estimate importances with totally randomized trees
330-
clf = ExtraTreesClassifier(n_estimators=500,
331-
max_features=1,
332-
criterion="entropy",
333-
random_state=0,
334-
presort=presort).fit(X, y)
324+
# Estimate importances with totally randomized trees
325+
clf = ExtraTreesClassifier(n_estimators=500,
326+
max_features=1,
327+
criterion="entropy",
328+
random_state=0).fit(X, y)
335329

336-
importances = sum(tree.tree_.compute_feature_importances(normalize=False)
337-
for tree in clf.estimators_) / clf.n_estimators
330+
importances = sum(tree.tree_.compute_feature_importances(normalize=False)
331+
for tree in clf.estimators_) / clf.n_estimators
338332

339-
# Check correctness
340-
assert_almost_equal(entropy(y), sum(importances))
341-
assert_less(np.abs(true_importances - importances).mean(), 0.01)
333+
# Check correctness
334+
assert_almost_equal(entropy(y), sum(importances))
335+
assert_less(np.abs(true_importances - importances).mean(), 0.01)
342336

343337

344338
def check_unfitted_feature_importances(name):
@@ -488,23 +482,21 @@ def check_multioutput(name):
488482
X_test = [[-1, -1], [1, 1], [-1, 1], [1, -1]]
489483
y_test = [[-1, 0], [1, 1], [-1, 2], [1, 3]]
490484

491-
for presort in True, False:
492-
est = FOREST_ESTIMATORS[name](random_state=0, bootstrap=False,
493-
presort=presort)
494-
y_pred = est.fit(X_train, y_train).predict(X_test)
495-
assert_array_almost_equal(y_pred, y_test)
496-
497-
if name in FOREST_CLASSIFIERS:
498-
with np.errstate(divide="ignore"):
499-
proba = est.predict_proba(X_test)
500-
assert_equal(len(proba), 2)
501-
assert_equal(proba[0].shape, (4, 2))
502-
assert_equal(proba[1].shape, (4, 4))
503-
504-
log_proba = est.predict_log_proba(X_test)
505-
assert_equal(len(log_proba), 2)
506-
assert_equal(log_proba[0].shape, (4, 2))
507-
assert_equal(log_proba[1].shape, (4, 4))
485+
est = FOREST_ESTIMATORS[name](random_state=0, bootstrap=False)
486+
y_pred = est.fit(X_train, y_train).predict(X_test)
487+
assert_array_almost_equal(y_pred, y_test)
488+
489+
if name in FOREST_CLASSIFIERS:
490+
with np.errstate(divide="ignore"):
491+
proba = est.predict_proba(X_test)
492+
assert_equal(len(proba), 2)
493+
assert_equal(proba[0].shape, (4, 2))
494+
assert_equal(proba[1].shape, (4, 4))
495+
496+
log_proba = est.predict_log_proba(X_test)
497+
assert_equal(len(log_proba), 2)
498+
assert_equal(log_proba[0].shape, (4, 2))
499+
assert_equal(log_proba[1].shape, (4, 4))
508500

509501

510502
def test_multioutput():

0 commit comments

Comments
 (0)