def test_hoeffding_tree_coverage():
max_samples = 1000
max_size_mb = 2
stream = RegressionGenerator(
n_samples=max_samples, n_features=10, n_informative=7, n_targets=3,
random_state=42
)
X, y = stream.next_sample(max_samples)
# Will generate a warning concerning the invalid leaf prediction option
tree = StackedSingleTargetHoeffdingTreeRegressor(
leaf_prediction='mean', grace_period=200,
memory_estimate_period=100, max_byte_size=max_size_mb*2**20
)
# Trying to predict without fitting
tree.predict(X[0])
tree.partial_fit(X, y)
# A tree without memory management enabled reaches over 3 MB in size
assert calculate_object_size(tree, 'MB') <= max_size_mb
tree = StackedSingleTargetHoeffdingTreeRegressor(
leaf_prediction='adaptive', grace_period=200,
memory_estimate_period=100, max_byte_size=max_size_mb*2**20,
learning_ratio_const=False
)
tree.partial_fit(X, y)
assert calculate_object_size(tree, 'MB') <= max_size_mb
def train_ssth():
print("Start training Stacked Single Target Hoeffding Tree Regressor...")
ssth_regressor = StackedSingleTargetHoeffdingTreeRegressor(
max_byte_size=2000000000)
start_time = time.time()
ssth_regressor.fit(X_train, y_train)
print("Execution ssth completed in -- %s seconds --" %
round(time.time() - start_time, 2))
pickle.dump(ssth_regressor, open("models/ssth_regressor.p", "wb"))
def test_hoeffding_tree_coverage(test_path):
# Cover nominal attribute observer
test_file = os.path.join(test_path, 'multi_target_regression_data.npz')
data = np.load(test_file)
X = data['X']
Y = data['Y']
# Will generate a warning concerning the invalid leaf prediction option
learner = StackedSingleTargetHoeffdingTreeRegressor(
leaf_prediction='mean',
nominal_attributes=[i for i in range(3)],
learning_ratio_const=False)
learner.partial_fit(X, Y)
def test_stacked_single_target_hoeffding_tree_regressor_adaptive(test_path):
stream = RegressionGenerator(n_samples=2000, n_features=20,
n_informative=15, random_state=1,
n_targets=3)
learner = StackedSingleTargetHoeffdingTreeRegressor(
leaf_prediction='adaptive',
random_state=1
)
cnt = 0
max_samples = 2000
wait_samples = 200
y_pred = np.zeros((int(max_samples / wait_samples), 3))
y_true = np.zeros((int(max_samples / wait_samples), 3))
while cnt < max_samples:
X, y = stream.next_sample()
# Test every n samples
if (cnt % wait_samples == 0) and (cnt != 0):
y_pred[int(cnt / wait_samples), :] = learner.predict(X)
y_true[int(cnt / wait_samples), :] = y
learner.partial_fit(X, y)
cnt += 1
test_file = os.path.join(
test_path,
'expected_preds_stacked_single_target_hoeffding_tree_adaptive.npy'
)
expected_predictions = np.load(test_file)
assert np.allclose(y_pred, expected_predictions)
error = mean_absolute_error(y_true, y_pred)
expected_error = 152.8716829154756
assert np.isclose(error, expected_error)
expected_info = "StackedSingleTargetHoeffdingTreeRegressor(binary_split=False, grace_period=200,\n" \
" leaf_prediction='adaptive',\n" \
" learning_ratio_const=True,\n" \
" learning_ratio_decay=0.001,\n" \
" learning_ratio_perceptron=0.02,\n" \
" max_byte_size=33554432,\n" \
" memory_estimate_period=1000000,\n" \
" nb_threshold=0, no_preprune=False,\n" \
" nominal_attributes=None,\n" \
" random_state=1,\n" \
" remove_poor_atts=False,\n" \
" split_confidence=1e-07,\n" \
" stop_mem_management=False,\n" \
" tie_threshold=0.05)"
assert learner.get_info() == expected_info
assert isinstance(learner.get_model_description(), type(''))
def test_stacked_single_target_hoeffding_tree_categorical_features(test_path):
data_path = os.path.join(test_path, 'ht_categorical_features_testcase.npy')
stream = np.load(data_path)
X, y = stream[:, :-2], stream[:, -2:]
nominal_attr_idx = np.arange(8)
learner = StackedSingleTargetHoeffdingTreeRegressor(
nominal_attributes=nominal_attr_idx, leaf_prediction='perceptron')
learner.partial_fit(X, y)
expected_description = "if Attribute 0 = -15.0:\n" \
" if Attribute 3 = 0.0:\n" \
" Leaf = Statistics {0: 80.0000, 1: [-192.4417, 80.0908], 2: [464.1268, 80.1882]}\n" \
" if Attribute 3 = 1.0:\n" \
" Leaf = Statistics {0: 77.0000, 1: [-184.8333, -7.2503], 2: [444.9068, 42.7423]}\n" \
" if Attribute 3 = 2.0:\n" \
" Leaf = Statistics {0: 56.0000, 1: [-134.1829, -1.0863], 2: [322.1336, 28.1218]}\n" \
" if Attribute 3 = 3.0:\n" \
" Leaf = Statistics {0: 62.0000, 1: [-148.2397, -17.2837], 2: [355.5327, 38.6913]}\n" \
"if Attribute 0 = 0.0:\n" \
" Leaf = Statistics {0: 672.0000, 1: [390.6773, 672.0472], 2: [761.0744, 672.1686]}\n" \
"if Attribute 0 = 1.0:\n" \
" Leaf = Statistics {0: 644.0000, 1: [671.3479, 174.3011], 2: [927.5194, 466.7064]}\n" \
"if Attribute 0 = 2.0:\n" \
" Leaf = Statistics {0: 619.0000, 1: [867.2865, 320.6506], 2: [1262.0992, 435.2835]}\n" \
"if Attribute 0 = 3.0:\n" \
" Leaf = Statistics {0: 627.0000, 1: [987.0864, 331.0822], 2: [1583.8108, 484.0456]}\n" \
"if Attribute 0 = -30.0:\n" \
" Leaf = Statistics {0: 88.0000, 1: [-269.7967, 25.9328], 2: [828.2289, 57.6501]}\n"
assert SequenceMatcher(None, expected_description,
learner.get_model_description()).ratio() > 0.9
new_sample = X[0].copy()
# Adding a new category on the split feature
new_sample[0] = 7
learner.predict([new_sample])
# Let's do the same considering other prediction strategy
learner = StackedSingleTargetHoeffdingTreeRegressor(
nominal_attributes=nominal_attr_idx, leaf_prediction='adaptive')
learner.partial_fit(X, y)
learner.predict([new_sample])
