def test_parameter_validation():
"""Test parameter validation checks for consistent assignment."""
test_cases = [
dict(
prediction_depth="nmlnp",
stopping_criteria=None,
),
dict(
prediction_depth="nmlnp",
stopping_criteria="not_a_float_or_a_callable",
),
dict(
prediction_depth="mlnp",
stopping_criteria=123.4,
),
dict(prediction_depth="some_invalid_prediction_depth_value", ),
dict(
algorithm="lcn",
training_strategy=None,
),
dict(
algorithm="lcn",
training_strategy="some_invalid_training_strategy",
),
dict(
algorithm="lcpn",
training_strategy="exclusive",
),
dict(algorithm="some_invalid_algorithm_value", ),
]
for classifier_kwargs in test_cases:
clf, (X, y) = make_classifier_and_data(**classifier_kwargs)
assert_that(calling(clf.fit).with_args(X=X, y=y), raises(TypeError))
def test_trivial_hierarchy_classification():
"""Test that a trivial (degenerate) hierarchy behaves as expected."""
clf, (X, y) = make_classifier_and_data(n_classes=5)
X_train, X_test, y_train, y_test = train_test_split(
X,
y,
test_size=0.30,
random_state=RANDOM_STATE,
)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
assert_that(accuracy, is_(close_to(1., delta=0.05)))
def test_fitted_attributes():
"""Test classifier attributes are set correctly after fitting."""
n_classes = 10
clf, (X, y) = make_classifier_and_data(n_classes=n_classes)
clf.fit(X, y)
assert_that(DiGraph(clf.class_hierarchy_),
matches_graph(DiGraph(clf.class_hierarchy)))
assert_that(clf.graph_, matches_graph(DiGraph(clf.class_hierarchy)))
assert_that(clf.classes_, contains_inanyorder(*range(n_classes)))
assert_that(clf.n_classes_, is_(equal_to(n_classes)))
assert_that(
clf.graph_.nodes[ROOT],
has_entries(metafeatures=has_entries(
n_samples=X.shape[0],
n_targets=n_classes,
), ),
)