def gender_performance(df_test_encoded, predictions, print_=False):
predictions_m = []
predictions_f = []
df_test_encoded_m = []
df_test_encoded_f = []
tamanho = len(df_test_encoded['sex'])
for i in range(tamanho):
if df_test_encoded['sex'].iloc[i] == 1:
df_test_encoded_m.append(df_test_encoded['earnings'].iloc[i])
predictions_m.append(predictions[i])
else:
df_test_encoded_f.append(df_test_encoded['earnings'].iloc[i])
predictions_f.append(predictions[i])
true_positive_m = true_positive_rate(df_test_encoded_m, predictions_m, pos_label=1)
false_positive_m = false_positive_rate(df_test_encoded_m, predictions_m, pos_label=1)
true_negative_m = true_negative_rate(df_test_encoded_m, predictions_m, pos_label=1)
false_negative_m = false_negative_rate(df_test_encoded_m, predictions_m, pos_label=1)
true_positive_f = true_positive_rate(df_test_encoded_f, predictions_f, pos_label=1)
false_positive_f = false_positive_rate(df_test_encoded_f, predictions_f, pos_label=1)
true_negative_f = true_negative_rate(df_test_encoded_f, predictions_f, pos_label=1)
false_negative_f = false_negative_rate(df_test_encoded_f, predictions_f, pos_label=1)
if print_:
print("True Positive Rate for Male:", true_positive_m)
print("True Positive Rate for Female:", true_positive_f)
print("False Positive Rate for Male:", false_positive_m)
print("False Positive Rate for Female:", false_positive_f)
print("True Negative Rate for Male:", true_negative_m)
print("True Negative Rate for Female:", true_negative_f)
print("False Negative Rate for Male:", false_negative_m)
print("False Negative Rate for Female:", false_negative_f)
def test_against_sklearn_weighted(self):
y_true = [
0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1, 0, 1,
0, 0, 1, 0
]
y_pred = [
0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1,
1, 0, 0, 0
]
weight = [
1, 2, 1, 1, 1, 3, 1, 4, 1, 2, 3, 4, 2, 3, 1, 2, 3, 1, 3, 2, 4, 2,
3, 1, 1, 5
]
actual = metrics.false_positive_rate(y_true,
y_pred,
sample_weight=weight)
tn, fp, fn, tp = skm.confusion_matrix(y_true,
y_pred,
sample_weight=weight).ravel()
assert (fp / (fp + tn)) == actual
actual = metrics.false_positive_rate(y_true,
y_pred,
pos_label=0,
sample_weight=weight)
tn, fp, fn, tp = skm.confusion_matrix(y_true,
y_pred,
labels=[1, 0],
sample_weight=weight).ravel()
assert (fp / (fp + tn)) == actual
def test_against_sklearn(self):
y_true = [0, 1, 0, 1, 0, 1, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 0, 1, 0, 0, 1, 0]
y_pred = [0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 0, 0]
actual = metrics.false_positive_rate(y_true, y_pred)
tn, fp, fn, tp = skm.confusion_matrix(y_true, y_pred).ravel()
assert (fp/(fp+tn)) == actual
actual = metrics.false_positive_rate(y_true, y_pred, pos_label=0)
tn, fp, fn, tp = skm.confusion_matrix(y_true, y_pred, labels=[1, 0]).ravel()
assert (fp/(fp+tn)) == actual
def test_some_correct(self):
y_true = [-1, -1, -1, -1, 1, 1, 1]
y_pred = [-1, 1, 1, -1, -1, 1, -1]
result = metrics.false_positive_rate(y_true, y_pred)
assert result == 0.5
result = metrics.false_positive_rate(y_true, y_pred, pos_label=1)
assert result == 0.5
result = metrics.false_positive_rate(y_true, y_pred, pos_label=-1)
assert result == pytest.approx(0.6666667)
def test_fpr_some_correct():
y_true = [0, 0, 0, 0, 1, 1, 1]
y_pred = [0, 1, 1, 0, 0, 1, 0]
result = metrics.false_positive_rate(y_true, y_pred)
assert result == 0.5
def test_all_negative_ones(self):
neg_ones = -1 * np.ones(10)
# Should behave as if pos_label=1
assert metrics.true_positive_rate(neg_ones, neg_ones) == 0
assert metrics.false_positive_rate(neg_ones, neg_ones) == 0
assert metrics.true_negative_rate(neg_ones, neg_ones) == 1
assert metrics.false_negative_rate(neg_ones, neg_ones) == 0
assert metrics.true_positive_rate(neg_ones, neg_ones, pos_label=1) == 0
assert metrics.false_positive_rate(neg_ones, neg_ones, pos_label=1) == 0
assert metrics.true_negative_rate(neg_ones, neg_ones, pos_label=1) == 1
assert metrics.false_negative_rate(neg_ones, neg_ones, pos_label=1) == 0
assert metrics.true_positive_rate(neg_ones, neg_ones, pos_label=-1) == 1
assert metrics.false_positive_rate(neg_ones, neg_ones, pos_label=-1) == 0
assert metrics.true_negative_rate(neg_ones, neg_ones, pos_label=-1) == 0
assert metrics.false_negative_rate(neg_ones, neg_ones, pos_label=-1) == 0
def test_all_ones(self):
ones = np.ones(10)
# should behave as if pos_label=1
assert metrics.true_positive_rate(ones, ones) == 1
assert metrics.false_positive_rate(ones, ones) == 0
assert metrics.true_negative_rate(ones, ones) == 0
assert metrics.false_negative_rate(ones, ones) == 0
assert metrics.true_positive_rate(ones, ones, pos_label=1) == 1
assert metrics.false_positive_rate(ones, ones, pos_label=1) == 0
assert metrics.true_negative_rate(ones, ones, pos_label=1) == 0
assert metrics.false_negative_rate(ones, ones, pos_label=1) == 0
assert metrics.true_positive_rate(ones, ones, pos_label=0) == 0
assert metrics.false_positive_rate(ones, ones, pos_label=0) == 0
assert metrics.true_negative_rate(ones, ones, pos_label=0) == 1
assert metrics.false_negative_rate(ones, ones, pos_label=0) == 0
def test_all_zeros(self):
zeros = np.zeros(10)
# Should behave as if pos_label=1
assert metrics.true_positive_rate(zeros, zeros) == 0
assert metrics.false_positive_rate(zeros, zeros) == 0
assert metrics.true_negative_rate(zeros, zeros) == 1
assert metrics.false_negative_rate(zeros, zeros) == 0
assert metrics.true_positive_rate(zeros, zeros, pos_label=1) == 0
assert metrics.false_positive_rate(zeros, zeros, pos_label=1) == 0
assert metrics.true_negative_rate(zeros, zeros, pos_label=1) == 1
assert metrics.false_negative_rate(zeros, zeros, pos_label=1) == 0
assert metrics.true_positive_rate(zeros, zeros, pos_label=0) == 1
assert metrics.false_positive_rate(zeros, zeros, pos_label=0) == 0
assert metrics.true_negative_rate(zeros, zeros, pos_label=0) == 0
assert metrics.false_negative_rate(zeros, zeros, pos_label=0) == 0
def test_none_correct(self):
y_true = [0, 0, 0, 0, 1]
y_pred = [1, 1, 1, 1, 0]
result = metrics.false_positive_rate(y_true, y_pred)
assert result == 1
