def test_fit(self):
ranker = PercentileRankOneFeature(feature="x3")
assert ranker.feature_importances_ is None
ranker.fit(x=self.data["X_train"], y=self.data["y_train"])
np.testing.assert_array_equal(
ranker.feature_importances_, np.array([0, 0, 1, 0])
)
def test_predict_proba(self):
for direction_value in [True, False]:
ranker = PercentileRankOneFeature(
feature="x3", low_value_high_score=direction_value)
ranker.fit(x=self.data["X_train"], y=self.data["y_train"])
results = ranker.predict_proba(self.data["X_test"])
if direction_value:
expected_ranks = [6, 1, 3, 0, 5, 2, 4, 5]
else:
expected_ranks = [0, 5, 3, 6, 1, 4, 2, 1]
assert scores_align_with_ranks(expected_ranks, results[:, 1])
def test_predict_proba(self):
for descend_value in [True, False]:
ranker = PercentileRankOneFeature(feature="x3",
descend=descend_value)
ranker.fit(x=self.data["X_train"], y=self.data["y_train"])
results = ranker.predict_proba(self.data["X_test"])
if descend_value:
expected_results = np.array([
np.zeros(len(self.data["X_test"])),
[0.875, 0.125, 0.375, 0, 0.625, 0.25, 0.5, 0.625],
]).transpose()
else:
expected_results = np.array([
np.zeros(len(self.data["X_test"])),
[0, 0.75, 0.5, 0.875, 0.125, 0.625, 0.375, 0.125],
]).transpose()
np.testing.assert_array_equal(results, expected_results)
def test_predict_proba(self):
for descend_value in [True, False]:
ranker = PercentileRankOneFeature(feature='x3',
descend=descend_value)
ranker.fit(x=self.data['X_train'], y=self.data['y_train'])
results = ranker.predict_proba(self.data['X_test'])
if descend_value:
expected_results = np.array([
np.zeros(len(self.data['X_test'])),
[.875, .125, .375, 0, .625, .25, .5, .625]
]).transpose()
else:
expected_results = np.array([
np.zeros(len(self.data['X_test'])),
[0, .75, .5, .875, .125, .625, .375, .125]
]).transpose()
np.testing.assert_array_equal(results, expected_results)
def test_ranking_on_unavailable_feature_raises_error(self):
ranker = PercentileRankOneFeature(feature="x5")
with self.assertRaises(BaselineFeatureNotInMatrix):
ranker.fit(x=self.data["X_train"], y=self.data["y_train"])
def test_fit(self):
ranker = PercentileRankOneFeature(feature='x3')
assert ranker.feature_importances_ == None
ranker.fit(x=self.data['X_train'], y=self.data['y_train'])
np.testing.assert_array_equal(ranker.feature_importances_,
np.array([0, 0, 1, 0]))