1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
|
import pytest
from sklearn.base import ClassifierMixin, clone
from sklearn.compose import make_column_transformer
from sklearn.datasets import load_iris
from sklearn.exceptions import NotFittedError
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import (
DetCurveDisplay,
PrecisionRecallDisplay,
RocCurveDisplay,
)
@pytest.fixture(scope="module")
def data():
return load_iris(return_X_y=True)
@pytest.fixture(scope="module")
def data_binary(data):
X, y = data
return X[y < 2], y[y < 2]
@pytest.mark.parametrize(
"Display", [DetCurveDisplay, PrecisionRecallDisplay, RocCurveDisplay]
)
def test_display_curve_error_non_binary(pyplot, data, Display):
"""Check that a proper error is raised when only binary classification is
supported."""
X, y = data
clf = DecisionTreeClassifier().fit(X, y)
msg = (
"Expected 'estimator' to be a binary classifier, but got DecisionTreeClassifier"
)
with pytest.raises(ValueError, match=msg):
Display.from_estimator(clf, X, y)
@pytest.mark.parametrize(
"response_method, msg",
[
(
"predict_proba",
"response method predict_proba is not defined in MyClassifier",
),
(
"decision_function",
"response method decision_function is not defined in MyClassifier",
),
(
"auto",
"response method decision_function or predict_proba is not "
"defined in MyClassifier",
),
(
"bad_method",
"response_method must be 'predict_proba', 'decision_function' or 'auto'",
),
],
)
@pytest.mark.parametrize(
"Display", [DetCurveDisplay, PrecisionRecallDisplay, RocCurveDisplay]
)
def test_display_curve_error_no_response(
pyplot,
data_binary,
response_method,
msg,
Display,
):
"""Check that a proper error is raised when the response method requested
is not defined for the given trained classifier."""
X, y = data_binary
class MyClassifier(ClassifierMixin):
def fit(self, X, y):
self.classes_ = [0, 1]
return self
clf = MyClassifier().fit(X, y)
with pytest.raises(ValueError, match=msg):
Display.from_estimator(clf, X, y, response_method=response_method)
@pytest.mark.parametrize(
"Display", [DetCurveDisplay, PrecisionRecallDisplay, RocCurveDisplay]
)
@pytest.mark.parametrize("constructor_name", ["from_estimator", "from_predictions"])
def test_display_curve_estimator_name_multiple_calls(
pyplot,
data_binary,
Display,
constructor_name,
):
"""Check that passing `name` when calling `plot` will overwrite the original name
in the legend."""
X, y = data_binary
clf_name = "my hand-crafted name"
clf = LogisticRegression().fit(X, y)
y_pred = clf.predict_proba(X)[:, 1]
# safe guard for the binary if/else construction
assert constructor_name in ("from_estimator", "from_predictions")
if constructor_name == "from_estimator":
disp = Display.from_estimator(clf, X, y, name=clf_name)
else:
disp = Display.from_predictions(y, y_pred, name=clf_name)
assert disp.estimator_name == clf_name
pyplot.close("all")
disp.plot()
assert clf_name in disp.line_.get_label()
pyplot.close("all")
clf_name = "another_name"
disp.plot(name=clf_name)
assert clf_name in disp.line_.get_label()
@pytest.mark.parametrize(
"clf",
[
LogisticRegression(),
make_pipeline(StandardScaler(), LogisticRegression()),
make_pipeline(
make_column_transformer((StandardScaler(), [0, 1])), LogisticRegression()
),
],
)
@pytest.mark.parametrize(
"Display", [DetCurveDisplay, PrecisionRecallDisplay, RocCurveDisplay]
)
def test_display_curve_not_fitted_errors(pyplot, data_binary, clf, Display):
"""Check that a proper error is raised when the classifier is not
fitted."""
X, y = data_binary
# clone since we parametrize the test and the classifier will be fitted
# when testing the second and subsequent plotting function
model = clone(clf)
with pytest.raises(NotFittedError):
Display.from_estimator(model, X, y)
model.fit(X, y)
disp = Display.from_estimator(model, X, y)
assert model.__class__.__name__ in disp.line_.get_label()
assert disp.estimator_name == model.__class__.__name__
|
