def test_mse_max_features():
sk = 0
iv = 0
n_times = 30
for X, y in _make_regression_datasets(n_times):
reg = skRegTree(max_depth=5,
max_features=4,
random_state=0)
reg.fit(X, y)
y_pred = reg.predict(X)
sk += np.mean((y - y_pred) ** 2)
reg = TreeRegressor(max_depth=5,
max_features=4,
random_state=0)
reg.fit(X, y)
y_pred = reg.predict(X)
iv += np.mean((y - y_pred) ** 2)
sk /= n_times
iv /= n_times
assert_almost_equal(sk, 4588.4, 1)
assert_almost_equal(iv, 4921.1, 1)
def test_mse_duplicate_features():
diabetes = load_diabetes()
X, y = diabetes.data, diabetes.target
reg = skRegTree(max_depth=5)
reg.fit(X, y)
y_pred = reg.predict(X)
sk = np.mean((y - y_pred)**2)
reg = TreeRegressor(max_depth=5)
reg.fit(X, y)
y_pred = reg.predict(X)
iv = np.mean((y - y_pred)**2)
assert_almost_equal(sk, iv)
def test_mse_duplicate_features():
diabetes = load_diabetes()
X, y = diabetes.data, diabetes.target
reg = skRegTree(max_depth=5)
reg.fit(X, y)
y_pred = reg.predict(X)
sk = np.mean((y - y_pred) ** 2)
reg = TreeRegressor(max_depth=5)
reg.fit(X, y)
y_pred = reg.predict(X)
iv = np.mean((y - y_pred) ** 2)
assert_almost_equal(sk, iv)
def test_mse_fully_developed():
sk = 0
iv = 0
for X, y in _make_regression_datasets(10):
reg = skRegTree(max_depth=None)
reg.fit(X, y)
y_pred = reg.predict(X)
sk += np.mean((y - y_pred)**2)
reg = TreeRegressor(max_depth=None)
reg.fit(X, y)
y_pred = reg.predict(X)
iv += np.mean((y - y_pred)**2)
assert_almost_equal(sk, iv)
def test_mse_fully_developed():
sk = 0
iv = 0
for X, y in _make_regression_datasets(10):
reg = skRegTree(max_depth=None)
reg.fit(X, y)
y_pred = reg.predict(X)
sk += np.mean((y - y_pred) ** 2)
reg = TreeRegressor(max_depth=None)
reg.fit(X, y)
y_pred = reg.predict(X)
iv += np.mean((y - y_pred) ** 2)
assert_almost_equal(sk, iv)
def test_mse_min_samples():
sk = 0
iv = 0
for X, y in _make_regression_datasets(10):
reg = skRegTree(max_depth=5, min_samples_split=4, min_samples_leaf=2)
reg.fit(X, y)
y_pred = reg.predict(X)
sk += np.mean((y - y_pred)**2)
reg = TreeRegressor(max_depth=5,
min_samples_split=4,
min_samples_leaf=2)
reg.fit(X, y)
y_pred = reg.predict(X)
iv += np.mean((y - y_pred)**2)
assert_almost_equal(sk, iv)
def test_mse_sample_weight():
sk = 0
iv = 0
n_times = 10
for X, y, w in _make_regression_datasets(n_times, sw=True):
reg = skRegTree(max_depth=5)
reg.fit(X, y, w)
y_pred = reg.predict(X)
sk += mean_squared_error(y, y_pred, sample_weight=w)
reg = TreeRegressor(max_depth=5)
reg.fit(X, y, w)
y_pred = reg.predict(X)
iv += mean_squared_error(y, y_pred, sample_weight=w)
sk /= n_times
iv /= n_times
assert_almost_equal(sk, iv)
def test_mse_min_samples():
sk = 0
iv = 0
for X, y in _make_regression_datasets(10):
reg = skRegTree(max_depth=5,
min_samples_split=4,
min_samples_leaf=2)
reg.fit(X, y)
y_pred = reg.predict(X)
sk += np.mean((y - y_pred) ** 2)
reg = TreeRegressor(max_depth=5,
min_samples_split=4,
min_samples_leaf=2)
reg.fit(X, y)
y_pred = reg.predict(X)
iv += np.mean((y - y_pred) ** 2)
assert_almost_equal(sk, iv)
def test_mse_sample_weight():
sk = 0
iv = 0
n_times = 10
for X, y, w in _make_regression_datasets(n_times, sw=True):
reg = skRegTree(max_depth=5)
reg.fit(X, y, w)
y_pred = reg.predict(X)
sk += mean_squared_error(y, y_pred, sample_weight=w)
reg = TreeRegressor(max_depth=5)
reg.fit(X, y, w)
y_pred = reg.predict(X)
iv += mean_squared_error(y, y_pred, sample_weight=w)
sk /= n_times
iv /= n_times
assert_almost_equal(sk, iv)
def test_mse_max_features():
sk = 0
iv = 0
n_times = 30
for X, y in _make_regression_datasets(n_times):
reg = skRegTree(max_depth=5, max_features=4, random_state=0)
reg.fit(X, y)
y_pred = reg.predict(X)
sk += np.mean((y - y_pred)**2)
reg = TreeRegressor(max_depth=5, max_features=4, random_state=0)
reg.fit(X, y)
y_pred = reg.predict(X)
iv += np.mean((y - y_pred)**2)
sk /= n_times
iv /= n_times
assert_almost_equal(sk, 4588.4, 1)
assert_almost_equal(iv, 4921.1, 1)