def test_hoeffding_tree():
stream = RegressionGenerator(n_samples=500,
n_features=20,
n_informative=15,
random_state=1)
stream.prepare_for_use()
learner = HoeffdingAdaptiveTreeRegressor(leaf_prediction='mean',
random_state=1)
cnt = 0
max_samples = 500
y_pred = array('d')
y_true = array('d')
wait_samples = 10
while cnt < max_samples:
X, y = stream.next_sample()
# Test every n samples
if (cnt % wait_samples == 0) and (cnt != 0):
y_pred.append(learner.predict(X)[0])
y_true.append(y[0])
learner.partial_fit(X, y)
cnt += 1
expected_predictions = array('d', [
102.38946041769101, 55.6584574987656, 5.746076599168373,
17.11797209372667, 2.566888222752787, 9.188247802192826,
17.87894804676911, 15.940629626883966, 8.981172175448485,
13.152624115190092, 11.106058099429399, 6.473195313058236,
4.723621479590173, 13.825568609556493, 8.698873073880696,
1.6452441811010252, 5.123496188584294, 6.34387187194982,
5.9977733790395105, 6.874251577667707, 4.605348088338317,
8.20112636572672, 9.032631648758098, 4.428189978974459,
4.249801041367518, 9.983272668044492, 12.859518508979734,
11.741395774380285, 11.230028410261868, 9.126921979081521,
9.132146661688296, 7.750655625124709, 6.445145118245414,
5.760928671876355, 4.041291302080659, 3.591837600560529,
0.7640424010500604, 0.1738639840537784, 2.2068337802212286,
-81.05302946841077, 96.17757415335177, -77.35894903819677,
95.85568683733698, 99.1981674250886, 99.89327888035015,
101.66673013734784, -79.1904234513751, -80.42952143783687,
100.63954789983896
])
assert np.allclose(y_pred, expected_predictions)
error = mean_absolute_error(y_true, y_pred)
expected_error = 143.11351404083086
assert np.isclose(error, expected_error)
expected_info = "HoeffdingAdaptiveTreeRegressor(binary_split=False, grace_period=200, leaf_prediction='mean', " \
"learning_ratio_const=True, learning_ratio_decay=0.001, learning_ratio_perceptron=0.02, " \
"max_byte_size=33554432, memory_estimate_period=1000000, nb_threshold=0, no_preprune=False, " \
"nominal_attributes=None, random_state=1, remove_poor_atts=False, split_confidence=1e-07, " \
"stop_mem_management=False, tie_threshold=0.05)"
info = " ".join([line.strip() for line in learner.get_info().split()])
assert info == expected_info
assert isinstance(learner.get_model_description(), type(''))
assert type(learner.predict(X)) == np.ndarray
def test_hoeffding_adaptive_tree_regressor_perceptron():
stream = RegressionGenerator(n_samples=500, n_features=20, n_informative=15, random_state=1)
learner = HoeffdingAdaptiveTreeRegressor(leaf_prediction='perceptron', random_state=1)
cnt = 0
max_samples = 500
y_pred = array('d')
y_true = array('d')
wait_samples = 10
while cnt < max_samples:
X, y = stream.next_sample()
# Test every n samples
if (cnt % wait_samples == 0) and (cnt != 0):
y_pred.append(learner.predict(X)[0])
y_true.append(y[0])
learner.partial_fit(X, y)
cnt += 1
expected_predictions = array('d', [207.20901655684412, 106.30316877540555, 101.46950096324191,
114.38162776688861, 48.40271620592212, -79.94375846313639,
-76.69182794940929, 88.38425569670662, -13.92372162581644,
3.0549887923350507, 55.36276732455883, 32.0512081208464,
17.54953203218902, -1.7305966738232161, 43.54548690756897,
8.502241407478213, -61.14739038895263, 50.528736810827745,
9.679668917948607, 89.93098085572623, 85.1994809437223,
1.8721866382932664, -7.1972581323107825, -45.86230662663542,
3.111671172363243, 57.921908276916646, 61.43400576850072,
-16.61695641848216, -6.0769944259948065, 19.929266442289546,
-60.972801351912224, -0.3342549973033524,
-50.53334350658139, -14.885488543743078,
-13.255920225124637, 28.909916365484275,
-103.03499425386107, -36.44921969674884, -15.40018796932204,
-84.98471039676006, 38.270205984888065, -62.97228157481581,
-48.095864628804044, 95.5028130171316, 73.62390886812497,
152.7135140597221, -120.4662342226783, -77.68182541723442,
66.82059046110074])
assert np.allclose(y_pred, expected_predictions)
error = mean_absolute_error(y_true, y_pred)
expected_error = 126.11208652969131
assert np.isclose(error, expected_error)
expected_info = "HoeffdingAdaptiveTreeRegressor(binary_split=False, grace_period=200, " \
"leaf_prediction='perceptron', learning_ratio_const=True, learning_ratio_decay=0.001, " \
"learning_ratio_perceptron=0.02, max_byte_size=33554432, memory_estimate_period=1000000, " \
"no_preprune=False, nominal_attributes=None, random_state=1, " \
"remove_poor_atts=False, split_confidence=1e-07, stop_mem_management=False, tie_threshold=0.05)"
info = " ".join([line.strip() for line in learner.get_info().split()])
assert info == expected_info
assert isinstance(learner.get_model_description(), type(''))
assert type(learner.predict(X)) == np.ndarray
assert learner._estimator_type == 'regressor'
def test_regression_hoeffding_adaptive_tree_categorical_features(test_path):
data_path = os.path.join(test_path, 'ht_categorical_features_testcase.npy')
stream = np.load(data_path)
# Removes the last column (used only in the multi-target regression case)
stream = stream[1000:, :-1]
X, y = stream[:, :-1], stream[:, -1]
nominal_attr_idx = np.arange(8)
# Typo in leaf prediction
learner = HoeffdingAdaptiveTreeRegressor(
nominal_attributes=nominal_attr_idx,
leaf_prediction='percptron'
)
learner.partial_fit(X, y)
expected_description = "if Attribute 1 = -1.0:\n" \
" if Attribute 0 = -15.0:\n" \
" Leaf = Statistics {0: 66.0000, 1: -164.9262, 2: 412.7679}\n" \
" if Attribute 0 = 0.0:\n" \
" Leaf = Statistics {0: 71.0000, 1: -70.3639, 2: 70.3179}\n" \
" if Attribute 0 = 1.0:\n" \
" Leaf = Statistics {0: 83.0000, 1: 0.9178, 2: 0.8395}\n" \
" if Attribute 0 = 2.0:\n" \
" Leaf = Statistics {0: 74.0000, 1: 73.6454, 2: 73.8353}\n" \
" if Attribute 0 = 3.0:\n" \
" Leaf = Statistics {0: 59.0000, 1: 75.2899, 2: 96.4856}\n" \
" if Attribute 0 = -30.0:\n" \
" Leaf = Statistics {0: 13.0000, 1: -40.6367, 2: 127.1607}\n" \
"if Attribute 1 = 0.0:\n" \
" if Attribute 0 = -15.0:\n" \
" Leaf = Statistics {0: 64.0000, 1: -158.0874, 2: 391.2359}\n" \
" if Attribute 0 = 0.0:\n" \
" Leaf = Statistics {0: 72.0000, 1: -0.4503, 2: 0.8424}\n" \
" if Attribute 0 = 1.0:\n" \
" Leaf = Statistics {0: 67.0000, 1: 68.0365, 2: 69.6664}\n" \
" if Attribute 0 = 2.0:\n" \
" Leaf = Statistics {0: 60.0000, 1: 77.7032, 2: 101.3210}\n" \
" if Attribute 0 = 3.0:\n" \
" Leaf = Statistics {0: 54.0000, 1: 77.4519, 2: 111.7702}\n" \
" if Attribute 0 = -30.0:\n" \
" Leaf = Statistics {0: 27.0000, 1: -83.8745, 2: 260.8891}\n" \
"if Attribute 1 = 1.0:\n" \
" Leaf = Statistics {0: 412.0000, 1: 180.7178, 2: 1143.9712}\n" \
"if Attribute 1 = 2.0:\n" \
" Leaf = Statistics {0: 384.0000, 1: 268.3498, 2: 1193.4180}\n" \
"if Attribute 1 = 3.0:\n" \
" Leaf = Statistics {0: 418.0000, 1: 289.5005, 2: 1450.7667}\n"
assert SequenceMatcher(
None, expected_description, learner.get_model_description()
).ratio() > 0.9
def test_regression_hoeffding_adaptive_tree_categorical_features(test_path):
data_path = os.path.join(test_path, 'ht_categorical_features_testcase.npy')
stream = np.load(data_path)
# Removes the last column (used only in the multi-target regression case)
stream = stream[1500:, :-1]
X, y = stream[:, :-1], stream[:, -1]
X = X[:, :-1]
nominal_attr_idx = np.arange(7)
# Typo in leaf prediction
learner = HoeffdingAdaptiveTreeRegressor(
nominal_attributes=nominal_attr_idx,
leaf_prediction='percptron',
random_state=1)
learner.partial_fit(X, y)
expected_description = "if Attribute 0 = -15.0:\n" \
" if Attribute 1 = -1.0:\n" \
" Leaf = Statistics {0: 37.0000, 1: -92.4231, 2: 231.1636}\n" \
" if Attribute 1 = 0.0:\n" \
" Leaf = Statistics {0: 38.0000, 1: -94.0931, 2: 233.4825}\n" \
" if Attribute 1 = 1.0:\n" \
" Leaf = Statistics {0: 55.0000, 1: -131.1069, 2: 312.9920}\n" \
" if Attribute 1 = 2.0:\n" \
" Leaf = Statistics {0: 38.0000, 1: -90.3821, 2: 215.5215}\n" \
" if Attribute 1 = 3.0:\n" \
" Leaf = Statistics {0: 54.0000, 1: -124.1223, 2: 285.7867}\n" \
"if Attribute 0 = 0.0:\n" \
" Leaf = Statistics {0: 123.0000, 1: 60.9178, 2: 132.5396}\n" \
"if Attribute 0 = 1.0:\n" \
" Leaf = Statistics {0: 124.0000, 1: 134.7770, 2: 184.4009}\n" \
"if Attribute 0 = 2.0:\n" \
" Leaf = Statistics {0: 104.0000, 1: 145.5842, 2: 212.8880}\n" \
"if Attribute 0 = 3.0:\n" \
" Leaf = Statistics {0: 118.0000, 1: 186.4441, 2: 300.6575}\n" \
"if Attribute 0 = -30.0:\n" \
" Leaf = Statistics {0: 88.0000, 1: -269.7967, 2: 828.2289}\n"
assert SequenceMatcher(None, expected_description,
learner.get_model_description()).ratio() > 0.9
def test_hoeffding_tree_perceptron():
stream = RegressionGenerator(n_samples=500,
n_features=20,
n_informative=15,
random_state=1)
stream.prepare_for_use()
learner = HoeffdingAdaptiveTreeRegressor(leaf_prediction='perceptron',
random_state=1)
cnt = 0
max_samples = 500
y_pred = array('d')
y_true = array('d')
wait_samples = 10
while cnt < max_samples:
X, y = stream.next_sample()
# Test every n samples
if (cnt % wait_samples == 0) and (cnt != 0):
y_pred.append(learner.predict(X)[0])
y_true.append(y[0])
learner.partial_fit(X, y)
cnt += 1
expected_predictions = array('d', [
1198.4326121743168, 456.36607750881586, 927.9912160545144,
1160.4797981899128, 506.50541829176535, -687.8187227095925,
-677.8120094065415, 231.14888704761225, -284.46324039942937,
-255.69195985557175, 47.58787439365423, -135.22494016284043,
-10.351457437330152, 164.95903200643997, 360.72854984472383,
193.30633911830088, -64.23638301570358, 587.9771578214296,
649.8395655757931, 481.01214222804026, 305.4402728117724,
266.2096493865043, -445.11447171009775, -567.5748694154349,
-68.70070048021438, -446.79910655850153, -115.892348067663,
-98.26862866231015, 71.04707905920286, -10.239274802165584,
18.748731569441812, 4.971217265129857, 172.2223575990573,
-655.2864976783711, -129.69921313686626, -114.01187375876822,
-405.66166686550963, -215.1264381928009, -345.91020370426247,
-80.49330468453074, 108.78958382083302, 134.95267043280126,
-398.5273538477553, -157.1784910649728, 219.72541225645654,
-100.91598162899217, 80.9768574308987, -296.8856956382453,
251.9332271253148
])
assert np.allclose(y_pred, expected_predictions)
error = mean_absolute_error(y_true, y_pred)
expected_error = 362.98595964244623
assert np.isclose(error, expected_error)
expected_info = "HoeffdingAdaptiveTreeRegressor(binary_split=False, grace_period=200, " \
"leaf_prediction='perceptron', learning_ratio_const=True, learning_ratio_decay=0.001, " \
"learning_ratio_perceptron=0.02, max_byte_size=33554432, memory_estimate_period=1000000, " \
"nb_threshold=0, no_preprune=False, nominal_attributes=None, random_state=1, " \
"remove_poor_atts=False, split_confidence=1e-07, stop_mem_management=False, tie_threshold=0.05)"
info = " ".join([line.strip() for line in learner.get_info().split()])
assert info == expected_info
assert isinstance(learner.get_model_description(), type(''))
assert type(learner.predict(X)) == np.ndarray
assert learner._estimator_type == 'regressor'
def test_hoeffding_adaptive_tree_regressor_alternate_tree():
learner = HoeffdingAdaptiveTreeRegressor(
leaf_prediction='mean', grace_period=1000, random_state=7
)
np.random.seed(8)
max_samples = 7000
cnt = 0
p1 = False
p2 = False
while cnt < max_samples:
X = [np.random.uniform(low=-1, high=1, size=2)]
if cnt < 3000:
if X[0][0] <= 0 and X[0][1] > 0:
y = [np.random.normal(loc=-3, scale=1)]
elif X[0][0] > 0 and X[0][1] > 0:
y = [np.random.normal(loc=3, scale=1)]
elif X[0][0] <= 0 and X[0][1] <= 0:
y = [np.random.normal(loc=3, scale=1)]
else:
y = [np.random.normal(loc=-3, scale=1)]
elif cnt < 5000:
if not p1:
expected_info = "if Attribute 0 <= 0.7308480624289246:\n" \
" if Attribute 1 <= 0.020068273107131107:\n" \
" Leaf = Statistics {0: 900.0000, 1: 685.4441, 2: 9052.7232}\n" \
" if Attribute 1 > 0.020068273107131107:\n" \
" Leaf = Statistics {0: 1716.0000, 1: -284.7812, 2: 17014.5944}\n" \
"if Attribute 0 > 0.7308480624289246:\n" \
" Leaf = Statistics {0: 384.0000, 1: -40.5676, 2: 3855.0453}\n"
assert expected_info == learner.get_model_description()
p1 = True
# Keep almost the same generation function
if X[0][0] <= 0 and X[0][1] > 0:
y = [np.random.normal(loc=-3, scale=1)]
elif X[0][0] > 0 and X[0][1] > 0:
y = [np.random.normal(loc=3, scale=1)]
elif X[0][0] <= 0 and X[0][1] <= 0:
y = [np.random.normal(loc=3, scale=1)]
else:
y = [np.random.normal(loc=-3, scale=1)]
# But shift the normal mean in a specific region
if X[0][0] <= 0.73:
y = [np.random.normal(loc=5, scale=0.1)]
elif cnt < 6000:
if not p2:
# Subtree swapped
expected_info = "if Attribute 0 <= 0.7308480624289246:\n" \
" if Attribute 0 <= 0.7210747610959465:\n" \
" Leaf = Statistics {0: 1447.0000, 1: 7229.8838, 2: 36138.9433}\n" \
" if Attribute 0 > 0.7210747610959465:\n" \
" Leaf = Statistics {0: 8.0000, 1: 30.5281, 2: 183.5354}\n" \
"if Attribute 0 > 0.7308480624289246:\n" \
" Leaf = Statistics {0: 654.0000, 1: -24.9928, 2: 6519.0335}\n"
assert expected_info == learner.get_model_description()
p2 = True
# Change how y is generated: only x_1 matters now
if X[0][1] > 0:
y = [np.random.normal(loc=20, scale=3)]
else:
y = [np.random.normal(loc=-20, scale=3)]
learner.partial_fit(X, y)
cnt += 1
# Root node changed
expected_info = "if Attribute 1 <= -0.00015267114158334927:\n" \
" Leaf = Statistics {0: 904.0000, 1: 1098.6423, 2: 332597.7050}\n" \
"if Attribute 1 > -0.00015267114158334927:\n" \
" Leaf = Statistics {0: 905.0000, 1: 17227.8522, 2: 332000.7548}\n"
assert expected_info == learner.get_model_description()
def test_hoeffding_adaptive_tree_regressor_alternate_tree():
learner = HoeffdingAdaptiveTreeRegressor(leaf_prediction='mean',
grace_period=1000,
random_state=7)
np.random.seed(8)
max_samples = 7000
cnt = 0
p1 = False
p2 = False
while cnt < max_samples:
X = [np.random.uniform(low=-1, high=1, size=2)]
if cnt < 3000:
if X[0][0] <= 0 and X[0][1] > 0:
y = [np.random.normal(loc=-3, scale=1)]
elif X[0][0] > 0 and X[0][1] > 0:
y = [np.random.normal(loc=3, scale=1)]
elif X[0][0] <= 0 and X[0][1] <= 0:
y = [np.random.normal(loc=3, scale=1)]
else:
y = [np.random.normal(loc=-3, scale=1)]
elif cnt < 5000:
if not p1:
expected_info = "if Attribute 0 <= 0.7308480624289246:\n" \
" if Attribute 1 <= 0.020068273107131107:\n" \
" Leaf = Statistics {0: 900.0000, 1: 685.4441, 2: 9052.7232}\n" \
" if Attribute 1 > 0.020068273107131107:\n" \
" Leaf = Statistics {0: 1716.0000, 1: -284.7812, 2: 17014.5944}\n" \
"if Attribute 0 > 0.7308480624289246:\n" \
" Leaf = Statistics {0: 384.0000, 1: -40.5676, 2: 3855.0453}\n"
model_description = learner.get_model_description()
assert expected_info == model_description
p1 = True
# Keep almost the same generation function
if X[0][0] <= 0 and X[0][1] > 0:
y = [np.random.normal(loc=-3, scale=1)]
elif X[0][0] > 0 and X[0][1] > 0:
y = [np.random.normal(loc=3, scale=1)]
elif X[0][0] <= 0 and X[0][1] <= 0:
y = [np.random.normal(loc=3, scale=1)]
else:
y = [np.random.normal(loc=-3, scale=1)]
# But shift the normal mean in a specific region
if X[0][0] <= 0.3:
y = [np.random.normal(loc=5, scale=0.1)]
elif cnt < 6000:
if not p2:
# Subtree swapped
expected_info = "if Attribute 0 <= 0.7308480624289246:\n" \
" if Attribute 0 <= 0.2979778083105622:\n" \
" Leaf = Statistics {0: 1108.0000, 1: 5539.5153, 2: 27706.8525}\n" \
" if Attribute 0 > 0.2979778083105622:\n" \
" Leaf = Statistics {0: 342.0000, 1: 48.2119, 2: 3518.3529}\n" \
"if Attribute 0 > 0.7308480624289246:\n" \
" Leaf = Statistics {0: 659.0000, 1: -28.8180, 2: 6546.5087}\n"
assert expected_info == learner.get_model_description()
p2 = True
# Change how y is generated: only x_1 matters now
if X[0][1] > 0:
y = [np.random.normal(loc=20, scale=3)]
else:
y = [np.random.normal(loc=-20, scale=3)]
learner.partial_fit(X, y)
cnt += 1
# Root node changed
expected_info = "if Attribute 1 <= 0.02469103490619995:\n" \
" Leaf = Statistics {0: 941.0000, 1: -18769.1383, 2: 378390.2088}\n" \
"if Attribute 1 > 0.02469103490619995:\n" \
" Leaf = Statistics {0: 900.0000, 1: -2030.2098, 2: 355715.9719}\n"
assert expected_info == learner.get_model_description()
def test_hoeffding_adaptive_tree_regressor_perceptron():
stream = RegressionGenerator(n_samples=500,
n_features=20,
n_informative=15,
random_state=1)
learner = HoeffdingAdaptiveTreeRegressor(leaf_prediction='perceptron',
random_state=1)
cnt = 0
max_samples = 500
y_pred = array('d')
y_true = array('d')
wait_samples = 10
while cnt < max_samples:
X, y = stream.next_sample()
# Test every n samples
if (cnt % wait_samples == 0) and (cnt != 0):
y_pred.append(learner.predict(X)[0])
y_true.append(y[0])
learner.partial_fit(X, y)
cnt += 1
expected_predictions = array('d', [
-106.84237763060068, -10.965517384802226, -180.90711470797237,
-218.20896751607663, -96.4271589961865, 110.51551963099622,
108.34616947202511, 30.1720109214627, 57.92205878998479,
77.82418885914053, 49.972060923364765, 68.56117081695875,
15.996949915551697, -34.22744443808294, -19.762696110319702,
-28.447329394752995, -50.62864370485592, -47.37357781048561,
-99.82613515424342, 13.985531117918336, 41.41709671929987,
-34.679807275938174, 62.75626094547859, 30.925078688018893,
12.130320819235365, 119.3648998377624, 82.96422756064737,
-6.920397563039609, -12.701774870569059, 24.883730398016034,
-74.22855883237567, -0.8012436194087567, -83.03683748750394,
46.737839617687854, 0.537404558240671, 48.53591837633138,
-86.2259777783834, -24.985514024179967, 6.396035456152859,
-90.19454995571908, 32.05821807667601, -83.08553684151566,
-28.32223999320023, 113.28916673506842, 68.10498750807977,
173.9146410394573, -150.2067507947196, -74.10346402222962,
54.39153137687993
])
assert np.allclose(y_pred, expected_predictions)
error = mean_absolute_error(y_true, y_pred)
expected_error = 115.78916175164417
assert np.isclose(error, expected_error)
expected_info = "HoeffdingAdaptiveTreeRegressor(binary_split=False, grace_period=200, " \
"leaf_prediction='perceptron', learning_ratio_const=True, learning_ratio_decay=0.001, " \
"learning_ratio_perceptron=0.02, max_byte_size=33554432, memory_estimate_period=1000000, " \
"no_preprune=False, nominal_attributes=None, random_state=1, " \
"remove_poor_atts=False, split_confidence=1e-07, stop_mem_management=False, tie_threshold=0.05)"
info = " ".join([line.strip() for line in learner.get_info().split()])
assert info == expected_info
assert isinstance(learner.get_model_description(), type(''))
assert type(learner.predict(X)) == np.ndarray
assert learner._estimator_type == 'regressor'