def test_classification_matrix_mismatch_in_class_names():
targets = ["A"] * 2 + ["B"] * 3
predictions = ["C"] * 2 + ["D"] * 3
matrix = ClassificationMatrix(targets, predictions)
assert matrix.tp("A") == 0
assert matrix.tn("A") == 0
assert matrix.fp("A") == 2
assert matrix.fn("A") == 0
assert matrix.tp("B") == 0
assert matrix.tn("B") == 0
assert matrix.fp("B") == 3
assert matrix.fn("B") == 0
assert matrix.tp("C") == 0
assert matrix.tn("C") == 0
assert matrix.fp("C") == 0
assert matrix.fn("C") == 2
assert matrix.tp("D") == 0
assert matrix.tn("D") == 0
assert matrix.fp("D") == 0
assert matrix.fn("D") == 3
assert matrix.accuracy() == 0
assert matrix.precision("A") == 0
assert matrix.recall("A") == 0
assert matrix.sensitivity("A") == 0
assert matrix.specificity("A") == 0
assert matrix.f1("A") == 0
assert matrix.matthews_corrcoef("A") == np.inf
assert matrix.precision("C") == 0
assert matrix.recall("C") == 0
assert matrix.sensitivity("C") == 0
assert matrix.specificity("C") == 0
assert matrix.f1("C") == 0
assert matrix.matthews_corrcoef("C") == np.inf
def test_classification_matrix_matthews_corrcoef():
targets = np.random.randint(2, size=20)
# Shuffle the middle part of the targets
section = targets[5:15]
np.random.shuffle(section)
predictions = np.r_[targets[:5], section, targets[15:]]
#
matrix = ClassificationMatrix(targets, predictions)
tp = matrix.tp()
fp = matrix.fp()
tn = matrix.tn()
fn = matrix.fn()
if (tp + fp) * (tp + fn) * (tn + fp) * (tn + fn) == 0:
assert matrix.matthews_corrcoef() == np.inf
else:
assert matrix.matthews_corrcoef() == (
(tp * tn - fp * fn) / ((tp + fp) * (tp + fn) * (tn + fp) * (tn + fn))
)
# Check the zero denominator
matrix = ClassificationMatrix([1, 1, 1, 0], [1, 1, 1, 1])
assert matrix.matthews_corrcoef() == np.inf