def setup(self):
(X_train, y_train), _ = get_test_data()
est = rgf.RegularizedGreedyForestClassifier(
l2=0.01, max_leaf_nodes=500)
est.fit(X_train, y_train)
self.ensemble = est.ensemble
def test_not_fitted_classifier(self):
(X_train, y_train), (X_test, y_test) = get_test_data()
est = rgf.RegularizedGreedyForestClassifier(
l2=0.01, max_leaf_nodes=500)
with pytest.raises(NotFittedError):
est.predict_proba(X_test)
with pytest.raises(NotFittedError):
est.predict(X_test)
def test_train_regression_regression(self):
(X_train, y_train), (X_test, y_test) = get_test_data(
classification=False)
est = rgf.RegularizedGreedyForestRegressor(
l2=0.01, max_leaf_nodes=500)
est.fit(X_train, y_train)
y_pred = est.predict(X_train)
fixture_name = get_fixture_path('rgf_regression_regression.npy')
saved_preds = np.loadtxt(fixture_name)
np.testing.assert_allclose(y_pred, saved_preds)
def test_train_regression(self):
(X_train, y_train), (X_test, y_test) = get_test_data(
classification=False)
est = rgf.RegularizedGreedyForestRegressor(
l2=0.01, max_leaf_nodes=500)
est.fit(X_train, y_train)
y_pred = est.predict(X_train)
train_score = metrics.mean_squared_error(y_train, y_pred)
assert train_score < 2
y_pred = est.predict(X_test)
test_score = metrics.mean_squared_error(y_test, y_pred)
assert test_score < 2
def test_classification_regression(self):
(X_train, y_train), (X_test, y_test) = get_test_data()
sample_weight = np.ones_like(y_train)
sample_weight[y_train == 0] = 0.5
est = rgf.RegularizedGreedyForestClassifier(
l2=0.01, max_leaf_nodes=500)
est.fit(X_train, y_train, sample_weight)
y_pred = est.predict(X_train)
fixture_name = get_fixture_path('rgf_classification_regression.npy')
saved_preds = np.loadtxt(fixture_name)
np.testing.assert_allclose(y_pred, saved_preds)
def test_train_regression_weighted(self):
(X_train, y_train), (X_test, y_test) = get_test_data(
classification=False)
sample_weight = np.ones_like(y_train)
sample_weight[y_train < 0] = 0.5
est = rgf.RegularizedGreedyForestRegressor(
l2=0.01, max_leaf_nodes=500)
est.fit(X_train, y_train, sample_weight)
y_pred = est.predict(X_train)
train_score = metrics.mean_squared_error(y_train, y_pred)
assert train_score < 2
y_pred = est.predict(X_test)
test_score = metrics.mean_squared_error(y_test, y_pred)
assert test_score < 2
def test_train_classification(self):
(X_train, y_train), (X_test, y_test) = get_test_data()
est = rgf.RegularizedGreedyForestClassifier(
l2=0.01, max_leaf_nodes=500)
est.fit(X_train, y_train)
y_pred = est.predict(X_train)
train_score = metrics.accuracy_score(y_train, y_pred)
assert train_score > 0.75
y_pred = est.predict(X_test)
test_score = metrics.accuracy_score(y_test, y_pred)
assert test_score > 0.75
y_proba = est.predict_proba(X_test)
test_score = metrics.roc_auc_score(y_test, y_proba)
assert test_score > 0.75
def test_pickle_classifier(self):
(X_train, y_train), (X_test, y_test) = get_test_data()
est = rgf.RegularizedGreedyForestClassifier(
l2=0.01, max_leaf_nodes=500)
est.fit(X_train, y_train)
y_pred = est.predict_proba(X_test)
with tempfile.NamedTemporaryFile('wr') as tfile:
file_name = tfile.name
try:
pickle.dump(est, open(file_name, 'wb'))
unpickle_est = pickle.load(open(file_name, 'rb'))
unpickle_y_pred = unpickle_est.predict_proba(X_test)
np.testing.assert_allclose(y_pred, unpickle_y_pred)
finally:
os.remove(file_name)
def test_train_classification_weighted(self):
(X_train, y_train), (X_test, y_test) = get_test_data()
sample_weight = np.ones_like(y_train)
sample_weight[y_train == 0] = 0.5
est = rgf.RegularizedGreedyForestClassifier(
l2=0.01, max_leaf_nodes=500)
est.fit(X_train, y_train, sample_weight)
y_pred = est.predict(X_train)
train_score = metrics.accuracy_score(y_train, y_pred)
assert train_score > 0.75
y_pred = est.predict(X_test)
test_score = metrics.accuracy_score(y_test, y_pred)
assert test_score > 0.75
y_proba = est.predict_proba(X_test)
test_score = metrics.roc_auc_score(y_test, y_proba)
assert test_score > 0.75
