def test_quantile_regression(self, boston_X, boston_y):
X_train, X_test, y_train, y_test = train_test_split(boston_X, boston_y)
forest = RangerForestRegressor(quantiles=False)
forest.fit(X_train, y_train)
assert not hasattr(forest, "random_node_values_")
with pytest.raises(ValueError):
forest.predict_quantiles(X_test, quantiles=[0.2, 0.5, 0.8])
forest = RangerForestRegressor(quantiles=True)
forest.fit(X_train, y_train)
assert hasattr(forest, "random_node_values_")
quantiles_lower = forest.predict_quantiles(X_test, quantiles=[0.1])
quantiles_upper = forest.predict_quantiles(X_test, quantiles=[0.9])
assert np.less(quantiles_lower, quantiles_upper).all()
assert quantiles_upper.ndim == 1
quantiles = forest.predict_quantiles(X_test, quantiles=[0.1, 0.9])
assert quantiles.shape == (X_test.shape[0], 2)
assert np.sum(np.isnan(quantiles_lower)) == 0
assert np.sum(np.isnan(quantiles_upper)) == 0
# test predict method
pred = forest.predict(X_test, quantiles=[0.2, 0.5])
assert pred.shape == (X_test.shape[0], 2)
assert np.sum(np.isnan(pred)) == 0
pred = forest.predict(X_test, quantiles=[0.2])
assert pred.ndim == 1
assert np.sum(np.isnan(pred)) == 0
# test with single record
boston_X_record = boston_X[0:1, :]
pred = forest.predict(boston_X_record, quantiles=[0.2, 0.5])
assert pred.shape == (1, 2)
assert np.sum(np.isnan(pred)) == 0
def test_predict(self, boston_X, boston_y):
forest = RangerForestRegressor()
forest.fit(boston_X, boston_y)
pred = forest.predict(boston_X)
assert len(pred) == boston_X.shape[0]
# test with single record
boston_X_record = boston_X[0:1, :]
pred = forest.predict(boston_X_record)
assert len(pred) == 1
def test_sample_fraction(self, boston_X, boston_y):
forest = RangerForestRegressor(sample_fraction=0.69)
forest.fit(boston_X, boston_y)
assert forest.sample_fraction_ == [0.69]
# test with single record
boston_X_record = boston_X[0:1, :]
pred = forest.predict(boston_X_record)
assert len(pred) == 1
def test_categorical_features(
self, boston_X, boston_y, respect_categorical_features
):
# add a categorical feature
categorical_col = np.atleast_2d(
np.array([random.choice([0, 1]) for _ in range(boston_X.shape[0])])
)
boston_X_c = np.hstack((boston_X, categorical_col.transpose()))
categorical_features = [boston_X.shape[1]]
forest = RangerForestRegressor(
respect_categorical_features=respect_categorical_features,
categorical_features=categorical_features,
)
if respect_categorical_features not in ["partition", "ignore", "order"]:
with pytest.raises(ValueError):
forest.fit(boston_X_c, boston_y)
return
forest.fit(boston_X_c, boston_y)
forest.predict(boston_X_c)
