def test_sklearn_compatibility_clf(self):
clf = DAGClassifier(
alpha=0.0,
fit_intercept=True,
dependent_target=True,
hidden_layer_units=[0],
standardize=True,
)
clf.get_params(deep=True)
def test_class_number_error(self, y):
clf = DAGClassifier(alpha=0.1)
X = pd.DataFrame(np.random.normal(size=(100, 2)))
with pytest.raises(
ValueError,
match="This solver needs samples of at least 2 classes"
" in the data, but the data contains only one"
" class: 0",
):
clf.fit(X, y)
def test_feature_importances_binary(self, hidden_layer_units):
clf = DAGClassifier(alpha=0.1, hidden_layer_units=hidden_layer_units)
X, y = (
pd.DataFrame(np.random.normal(size=(100, 2))),
pd.Series(np.zeros(shape=(100, ), dtype=int)),
)
y[X[0] < 0] = 1
clf.fit(X, y)
assert isinstance(clf.feature_importances_, np.ndarray)
coef_ = pd.DataFrame(clf.feature_importances_, columns=X.columns)
# assert that the sign of the coefficient is positive for both nonlinear and linear cases
assert coef_.iloc[0, 0] > 0
def test_wrong_target_dist_error(self, target_dist_type):
with pytest.raises(
NotImplementedError,
# match=f"Currently only implements [{', '.join(DAGClassifier._supported_types)}] dist types."
# " Got: {target_dist_type}",
):
DAGClassifier(target_dist_type=target_dist_type)
def test_coef_categorical(self, hidden_layer_units):
clf = DAGClassifier(alpha=0.1, hidden_layer_units=hidden_layer_units)
X, y = (
pd.DataFrame(np.random.normal(size=(100, 2))),
pd.Series(np.zeros(shape=(100, ), dtype=int)),
)
y[X[0] < -0.1] = 1
y[X[0] > 0.1] = 2
clf.fit(X, y)
assert isinstance(clf.coef_, np.ndarray)
assert clf.coef_.shape == (3, 2)
coef_ = pd.DataFrame(clf.coef_, columns=X.columns)
# second category is made likely by negative X
assert coef_.iloc[1, 0] < 0
# third category is made likely by positive X
assert coef_.iloc[2, 0] > 0
def test_value_predict_type_binary(self, y_type):
clf = DAGClassifier(alpha=0.1)
X, y = (
pd.DataFrame(np.random.normal(size=(100, 2))),
pd.Series(np.zeros(shape=(100, ), dtype=y_type)),
)
y[X[0] < 0] = y_type(1)
clf.fit(X, y)
y_pred = clf.predict(X)
assert isinstance(y_pred[0], y_type)
y_pred_proba = clf.predict_proba(X)
assert isinstance(y_pred_proba[0, 0], np.float64)
assert len(y_pred_proba.shape) == 2
def test_glm(self, target_dist_type, y):
clf = DAGClassifier(target_dist_type=target_dist_type)
X = np.random.normal(size=(100, 2))
clf.fit(X, y)
clf.predict(X)