def test_isoup_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 = iSOUPTreeRegressor(
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 = iSOUPTreeRegressor(
nominal_attributes=nominal_attr_idx,
leaf_prediction='adaptive'
)
learner.partial_fit(X, y)
learner.predict([new_sample])
def test_isoup_tree_model_description():
stream = RegressionGenerator(n_samples=700,
n_features=20,
n_informative=15,
random_state=1,
n_targets=3)
learner = iSOUPTreeRegressor(leaf_prediction='mean')
max_samples = 700
X, y = stream.next_sample(max_samples)
# Trying to predict without fitting
learner.predict(X[0])
learner.partial_fit(X, y)
expected_description = "if Attribute 11 <= 0.36737233297880056:\n" \
" Leaf = Statistics {0: 450.0000, 1: [-23322.8079, -30257.1616, -18740.9462], " \
"2: [22242706.1751, 29895648.2424, 18855571.7943]}\n" \
"if Attribute 11 > 0.36737233297880056:\n" \
" Leaf = Statistics {0: 250.0000, 1: [33354.8675, 32390.6094, 22886.4176], " \
"2: [15429435.6709, 17908472.4289, 10709746.1079]}\n" \
assert SequenceMatcher(None, expected_description,
learner.get_model_description()).ratio() > 0.9
def demo(output_file=None):
""" Test iSOUP-Tree
This demo demonstrates how to evaluate a iSOUP-Tree multi-target regressor.
Parameters
----------
output_file: string
The name of the csv output file
"""
stream = RegressionGenerator(n_samples=5000,
n_features=20,
n_informative=15,
random_state=1,
n_targets=7)
regressor = iSOUPTreeRegressor(leaf_prediction='adaptive')
# Setup the evaluator
evaluator = EvaluatePrequential(pretrain_size=1,
batch_size=1,
n_wait=200,
max_time=1000,
output_file=output_file,
show_plot=False,
metrics=[
'average_mean_square_error',
'average_mean_absolute_error',
'average_root_mean_square_error'
])
# Evaluate
evaluator.evaluate(stream=stream, model=regressor)
def test_evaluate_multi_target_regression_coverage(tmpdir):
from skmultiflow.data import RegressionGenerator
from skmultiflow.trees import iSOUPTreeRegressor
max_samples = 1000
# Stream
stream = RegressionGenerator(n_samples=max_samples,
n_features=20,
n_informative=15,
random_state=1,
n_targets=7)
# Learner
mtrht = iSOUPTreeRegressor(leaf_prediction='adaptive')
output_file = os.path.join(str(tmpdir), "prequential_summary.csv")
metrics = [
'average_mean_square_error', 'average_mean_absolute_error',
'average_root_mean_square_error'
]
evaluator = EvaluatePrequential(max_samples=max_samples,
metrics=metrics,
output_file=output_file)
evaluator.evaluate(stream=stream, model=mtrht, model_names=['MTRHT'])
def test_evaluate_delayed_multi_target_regression_coverage(tmpdir):
from skmultiflow.data import RegressionGenerator
from skmultiflow.trees import iSOUPTreeRegressor
max_samples = 1000
# Stream
data = RegressionGenerator(n_samples=max_samples,
n_features=20,
n_informative=15,
random_state=1,
n_targets=7)
# Get X and y
X, y = data.next_sample(max_samples)
time = generate_random_dates(seed=1, samples=max_samples)
# Setup temporal stream
stream = TemporalDataStream(X, y, time, ordered=False)
# Learner
mtrht = iSOUPTreeRegressor(leaf_prediction='adaptive')
output_file = os.path.join(str(tmpdir), "prequential_delayed_summary.csv")
metrics = [
'average_mean_square_error', 'average_mean_absolute_error',
'average_root_mean_square_error'
]
evaluator = EvaluatePrequentialDelayed(max_samples=max_samples,
metrics=metrics,
output_file=output_file)
evaluator.evaluate(stream=stream, model=mtrht, model_names=['MTRHT'])
def test_get_tags():
classifier = HoeffdingTreeClassifier()
regressor = HoeffdingTreeRegressor()
multi_output_regressor = iSOUPTreeRegressor()
classifier_tags = classifier._get_tags()
expected_tags = {
'X_types': ['2darray'],
'_skip_test': False,
'allow_nan': False,
'multilabel': False,
'multioutput': False,
'multioutput_only': False,
'no_validation': False,
'non_deterministic': False,
'poor_score': False,
'requires_positive_data': False,
'stateless': False
}
assert classifier_tags == expected_tags
regressor_tags = regressor._get_tags()
expected_tags = {
'X_types': ['2darray'],
'_skip_test': False,
'allow_nan': False,
'multilabel': False,
'multioutput': False,
'multioutput_only': False,
'no_validation': False,
'non_deterministic': False,
'poor_score': False,
'requires_positive_data': False,
'stateless': False
}
assert regressor_tags == expected_tags
multi_output_regressor_tags = multi_output_regressor._get_tags()
expected_tags = {
'X_types': ['2darray'],
'_skip_test': False,
'allow_nan': False,
'multilabel': False,
'multioutput': True,
'multioutput_only': True,
'no_validation': False,
'non_deterministic': False,
'poor_score': False,
'requires_positive_data': False,
'stateless': False
}
assert multi_output_regressor_tags == expected_tags
def test_isoup_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)
# Cover memory management
tree = iSOUPTreeRegressor(leaf_prediction='mean',
grace_period=200,
memory_estimate_period=100,
max_byte_size=max_size_mb * 2**20)
# Invalid split_criterion
tree.split_criterion = 'ICVR'
X, y = stream.next_sample(max_samples)
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
# Memory management in a tree with perceptron leaves (purposeful typo in leaf_prediction)
tree = iSOUPTreeRegressor(leaf_prediction='PERCEPTRON',
grace_period=200,
memory_estimate_period=100,
max_byte_size=max_size_mb * 2**20)
tree.partial_fit(X, y)
assert calculate_object_size(tree, 'MB') <= max_size_mb
# Memory management in a tree with adaptive leaves
tree = iSOUPTreeRegressor(leaf_prediction='adaptive',
grace_period=200,
memory_estimate_period=100,
max_byte_size=max_size_mb * 2**20)
tree.partial_fit(X, y)
assert calculate_object_size(tree, 'MB') <= max_size_mb
def test_isoup_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']
# Invalid leaf prediction option
learner = iSOUPTreeRegressor(leaf_prediction='MEAN',
nominal_attributes=[i for i in range(3)])
print(learner.split_criterion)
# Invalid split_criterion
learner.split_criterion = 'ICVR'
learner.partial_fit(X, Y)
def test_isoup_tree_mean(test_path):
stream = RegressionGenerator(n_samples=2000,
n_features=20,
n_informative=15,
random_state=1,
n_targets=3)
stream.prepare_for_use()
learner = iSOUPTreeRegressor(leaf_prediction='mean')
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_multi_target_regression_mean.npy')
expected_predictions = np.load(test_file)
assert np.allclose(y_pred, expected_predictions)
error = mean_absolute_error(y_true, y_pred)
expected_error = 191.2823924547882
assert np.isclose(error, expected_error)
expected_info = "iSOUPTreeRegressor(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=None, 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 type(learner.predict(X)) == np.ndarray
from skmultiflow.data import RegressionGenerator, MultilabelGenerator
from skmultiflow.trees import iSOUPTreeRegressor
# Setup a data stream
n_targets = 20
regression_stream = RegressionGenerator(n_targets=n_targets,
random_state=1,
n_samples=200)
ml_stream = MultilabelGenerator(random_state=1,
n_samples=200,
n_targets=n_targets,
n_features=10)
# Setup iSOUP Tree Regressor
isoup_tree = iSOUPTreeRegressor()
# Auxiliary variables to control loop and track performance
n_samples = 0
max_samples = 200
y_pred = np.zeros((max_samples, n_targets))
y_true = np.zeros((max_samples, n_targets))
# Run test-then-train loop for max_samples and while there is data
while n_samples < max_samples and ml_stream.has_more_samples():
X, y = ml_stream.next_sample()
y_true[n_samples] = y[0]
y_pred[n_samples] = isoup_tree.predict(X)[0]
isoup_tree.partial_fit(X, y)
n_samples += 1
)
},
"oza_ml_cc_ht": {
"name": "OzaBagging (cc) / ecc - ht",
"model": lambda data_stream: ClassifierChain(
HoeffdingTreeClassifier(),
n_targets=data_stream.n_targets
),
"ensemble": lambda model, stream: OzaBaggingMLClassifier(
base_estimator=model,
n_estimators=10
)
},
"isoup": {
"name": "iSoup-Tree",
"model": lambda _: iSOUPTreeRegressor(),
"ensemble": False
},
}
DEFAULT_DATASETS = ["enron", "mediamill", "20NG"]
parser = argparse.ArgumentParser("Script to classify streams")
parser.add_argument("-e", "--experiment", help="Description of the experiment")
parser.add_argument("-d", "--datasets",
help="List of skmultilearn datasets (including 20NG)",
nargs="*", default=DEFAULT_DATASETS)
parser.add_argument("-m", "--models", help="List of models to train data",
nargs='*', default=SUPPORTED_MODELS.keys())
parser.add_argument("-s", "--streams", help="Path to stream", nargs='*')
parser.add_argument("-S", "--streamsnames", help="Names of streams", nargs='*')
parser.add_argument("-l", "--labels", type=int, help="Number of labels")
