def test_mcc_inverse():
golds = np.concatenate(
(np.arange(2), np.random.randint(0, 2, size=np.random.randint(4,
100))),
axis=0)
preds = 1 - golds
cm = ConfusionMatrix.create(golds, preds)
np.testing.assert_allclose(cm.get_mcc(), -1.0, TOL)
def test():
C = np.random.randint(3, 11)
golds = np.concatenate(
(np.arange(C),
np.random.randint(0, C, size=np.random.randint(4, 100))),
axis=0)
preds = np.random.randint(0, C, size=len(golds))
cm = ConfusionMatrix.create(golds, preds)
np.testing.assert_allclose(cm.get_r_k(), explicit_r_k(cm._cm), TOL)
def test_r_k_inverse():
"""The worst value for R k ranges from -1 to 0 depending on the distribution of the true labels."""
C = np.random.randint(2, 11)
golds = np.concatenate(
(np.arange(C), np.random.randint(0, C, size=np.random.randint(4,
100))),
axis=0)
preds = golds + 1 % C
cm = ConfusionMatrix.create(golds, preds)
assert -1.0 <= cm.get_r_k() <= 0.0
def test_r_k_perfect():
"""Perfect correlation results in a score of 1."""
C = np.random.randint(2, 11)
golds = np.concatenate(
(np.arange(C), np.random.randint(0, C, size=np.random.randint(4,
100))),
axis=0)
preds = np.copy(golds)
cm = ConfusionMatrix.create(golds, preds)
np.testing.assert_allclose(cm.get_r_k(), 1.0)
def test():
golds = np.concatenate(
(np.arange(2),
np.random.randint(0, 2, size=np.random.randint(4, 100))),
axis=0)
preds = np.random.randint(0, 2, size=len(golds))
cm = ConfusionMatrix.create(golds, preds)
np.testing.assert_allclose(cm.get_mcc(), explicit_mcc(golds, preds),
TOL)
np.testing.assert_allclose(cm.get_mcc(), author_mcc(cm._cm), TOL, TOL)
def test_mcc_example():
Y_TRUE = [1, 1, 1, 0]
Y_PRED = [1, 0, 1, 1]
MCC_GOLD = -0.3333333333333333
cm = ConfusionMatrix.create(Y_TRUE, Y_PRED)
np.testing.assert_allclose(cm.get_mcc(), MCC_GOLD, TOL)
def test_create_cm():
gold = make_mc_cm()
cm = ConfusionMatrix.create(Y_TRUE, Y_PRED)
np.testing.assert_equal(gold._cm, cm._cm)