def test_oza_bagging():
stream = SEAGenerator(1, noise_percentage=0.067, random_state=112)
knn = KNNClassifier(n_neighbors=8, leaf_size=40, max_window_size=2000)
learner = OzaBaggingClassifier(base_estimator=knn,
n_estimators=3,
random_state=112)
first = True
cnt = 0
max_samples = 5000
predictions = []
wait_samples = 100
correct_predictions = 0
while cnt < max_samples:
X, y = stream.next_sample()
# Test every n samples
if (cnt % wait_samples == 0) and (cnt != 0):
predictions.append(learner.predict(X)[0])
if y[0] == predictions[-1]:
correct_predictions += 1
if first:
learner.partial_fit(X, y, classes=stream.target_values)
first = False
else:
learner.partial_fit(X, y)
cnt += 1
performance = correct_predictions / len(predictions)
expected_predictions = [
1, 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 1, 1, 0, 1, 0, 0, 1, 0, 0,
1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
1
]
assert np.alltrue(predictions == expected_predictions)
expected_performance = 0.8979591836734694
assert np.isclose(expected_performance, performance)
expected_correct_predictions = 44
assert correct_predictions == expected_correct_predictions
assert type(learner.predict(X)) == np.ndarray
assert type(learner.predict_proba(X)) == np.ndarray
expected_info = "OzaBaggingClassifier(base_estimator=KNNClassifier(leaf_size=40, " \
"max_window_size=2000, metric='euclidean', n_neighbors=8), " \
"n_estimators=3, random_state=112)"
info = " ".join([line.strip() for line in learner.get_info().split()])
assert info == expected_info
def demo():
""" _test_oza_bagging
This demo tests the OzaBaggingClassifier using KNNADWINClassifier as base estimator
on samples given by a SEAGenerator.
The test computes the performance of the OzaBaggingClassifier as well
as the time to create the structure and classify max_samples (5000 by
default) instances.
"""
logging.basicConfig(format='%(message)s', level=logging.INFO)
warnings.filterwarnings("ignore", ".*Passing 1d.*")
stream = SEAGenerator(1, noise_percentage=.067, random_state=1)
clf = OzaBaggingClassifier(base_estimator=KNNADWINClassifier(
n_neighbors=8, max_window_size=2000, leaf_size=30),
n_estimators=2,
random_state=1)
sample_count = 0
correctly_classified = 0
max_samples = 5000
train_size = 8
first = True
if train_size > 0:
X, y = stream.next_sample(train_size)
clf.partial_fit(X, y, classes=stream.target_values)
first = False
while sample_count < max_samples:
if sample_count % (max_samples / 20) == 0:
logging.info('%s%%', str((sample_count // (max_samples / 20) * 5)))
X, y = stream.next_sample()
my_pred = clf.predict(X)
if first:
clf.partial_fit(X, y, classes=stream.target_values)
first = False
else:
clf.partial_fit(X, y)
if my_pred is not None:
if y[0] == my_pred[0]:
correctly_classified += 1
sample_count += 1
print(str(sample_count) + ' samples analyzed.')
print('My performance: ' + str(correctly_classified / sample_count))
# Applying Oza Bagging Classifier on a synthetic data stream
from skmultiflow.meta import OzaBaggingClassifier
from skmultiflow.lazy import KNNADWINClassifier
from skmultiflow.evaluation import EvaluatePrequential
from skmultiflow.data.sea_generator import SEAGenerator
# Simulate the data stream
dstream = SEAGenerator(classification_function=2,
balance_classes=True,
noise_percentage=0.3,
random_state=333)
# Instantiate the Oza Bagging classifier method with KNN ADWIN classifier as the base model
oza_class = OzaBaggingClassifier(base_estimator=KNNADWINClassifier(
n_neighbors=10, max_window_size=1000),
n_estimators=6,
random_state=333)
# Prequential Evaluation
evaluate1 = EvaluatePrequential(show_plot=False,
pretrain_size=1000,
max_samples=10000,
metrics=['accuracy'])
# Run the evaluation
evaluate1.evaluate(stream=dstream, model=oza_class)
###################################################
# Applying Leveraging Bagging Classifier on a synthetic data stream
from skmultiflow.meta import LeveragingBaggingClassifier
from skmultiflow.lazy import KNNADWINClassifier
#data, X, y = read_data_csv('./data/streaming-datasets-master/elec.csv')
#data, X, y = read_data_csv('./data/streaming-datasets-master/airlines.csv')
#data, X, y = read_data_csv('./data/streaming-datasets-master/agr_a.csv')
#data, X, y = read_data_csv('./data/streaming-datasets-master/covtype.csv')
stream = DataStream(X, y)
stream.prepare_for_use()
# 2a. Models initialization
nb = NaiveBayes()
ht = HoeffdingTreeClassifier()
aw = AccuracyWeightedEnsembleClassifier()
dw = DynamicWeightedMajorityClassifier()
ob = OnlineBoostingClassifier()
oz = OzaBaggingClassifier()
# 2b. Inicialization of DDCW model for comparsion tests
dwc = DiversifiedDynamicClassWeightedClassifier(
period=100,
base_estimators=[NaiveBayes(), HoeffdingTreeClassifier()],
min_estimators=5,
max_estimators=20,
alpha=0.2,
beta=3,
theta=0.2) #0.5
# 2c. Inicialization of DDCW models for parameter testing
#ht1_p1 = DiversifiedDynamicClassWeightedClassifier(period=500)
#ht1_p5 = DiversifiedDynamicClassWeightedClassifier(period=500)
#ht1_p10 = DiversifiedDynamicClassWeightedClassifier(period=1000)