def test_equal_opp_edge_3(self):
# Data: homogeneous both groups in ground truth - returns nan
y_true = np.array([0, 0, 0, 0, 1, 1, 1, 1])
y_pred = np.array([0, 1, 1, 1, 1, 1, 1, 0])
is_member = np.array([1, 1, 1, 1, 0, 0, 0, 0])
# Metric
metric = BinaryFairnessMetrics.EqualOpportunity()
with self.assertWarns(
UserWarning): # division by zero caught inside numpy
metric.get_score(y_true, y_pred, is_member)
def test_equal_opp_normal_invalid(self):
# Data
y_true = np.array([1, 0, 0, 0, 1, 1, 0, 2])
y_pred = np.array([0, 1, 1, 1, 1, 1, 1, 0])
is_member = np.array([1, 1, 1, 1, 0, 0, 0, 0])
# Metric
metric = BinaryFairnessMetrics.EqualOpportunity()
# Score
with self.assertRaises(ValueError):
metric.get_score(y_true, y_pred, is_member)
def test_equal_opp_edge_2(self):
# Data
y_true = np.array([1, 0, 0, 0, 1, 1, 1, 1])
y_pred = np.array([1, 1, 1, 1, 0, 0, 0, 0])
# edge case equal opp == -1
is_member = np.array([0, 0, 0, 0, 1, 1, 1, 1])
# Metric
metric = BinaryFairnessMetrics.EqualOpportunity()
with self.assertWarns(
UserWarning): # division by zero caught inside numpy
assert metric.get_score(y_true, y_pred, is_member) == -1
def test_equal_opp_normal_list(self):
# Data
y_true = [1, 0, 0, 0, 1, 1, 0, 1]
y_pred = [0, 1, 1, 1, 1, 1, 1, 0]
is_member = [1, 1, 1, 1, 0, 0, 0, 0]
# Metric
metric = BinaryFairnessMetrics.EqualOpportunity()
# Score
score = metric.get_score(y_true, y_pred, is_member)
assert np.isclose(score, -0.666, atol=0.001)
def test_equal_opp_normal_df(self):
# medium number
my_df = pd.DataFrame.from_dict({
'y_true': [1, 0, 0, 0, 1, 1, 0, 1],
'y_pred': [0, 1, 1, 1, 1, 1, 1, 0],
'is_member': [1, 1, 1, 1, 0, 0, 0, 0]
})
# Metric
metric = BinaryFairnessMetrics.EqualOpportunity()
# Score
score = metric.get_score(my_df['y_true'], my_df['y_pred'],
my_df['is_member'])
assert np.isclose(score, -0.666, atol=0.001)
