def test_knn():
stream = SEAGenerator(random_state=1)
stream.prepare_for_use()
learner = KNN(n_neighbors=8, max_window_size=2000, leaf_size=40)
cnt = 0
max_samples = 5000
predictions = array('i')
correct_predictions = 0
wait_samples = 100
X_batch = []
y_batch = []
while cnt < max_samples:
X, y = stream.next_sample()
X_batch.append(X[0])
y_batch.append(y[0])
# 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
learner.partial_fit(X, y)
cnt += 1
expected_predictions = array('i', [
1, 1, 1, 0, 1, 1, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 0, 0, 1, 0,
1
])
assert np.alltrue(predictions == expected_predictions)
expected_correct_predictions = 49
assert correct_predictions == expected_correct_predictions
expected_info = 'KNN(leaf_size=40, max_window_size=2000, n_neighbors=8, nominal_attributes=None)'
assert learner.get_info() == expected_info
learner.reset()
assert learner.get_info() == expected_info
X_batch = np.array(X_batch)
y_batch = np.array(y_batch)
learner.fit(X_batch[:4500], y_batch[:4500], classes=[0, 1])
predictions = learner.predict(X_batch[4501:4550])
expected_predictions = array('i', [
1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 0, 1, 1,
1, 1, 0, 0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1,
0
])
assert np.alltrue(predictions == expected_predictions)
correct_predictions = sum(predictions == y_batch[4501:4550])
expected_correct_predictions = 49
assert correct_predictions == expected_correct_predictions
assert type(learner.predict(X)) == np.ndarray
assert type(learner.predict_proba(X)) == np.ndarray
def hyperparametertuning_classifiers(learn, X, y, knn_max_w_size):
cl_name = learn.__class__.__name__
# print (cl_name)
scor = 'balanced_accuracy'
cv = 10
if cl_name == 'KNN':
KNN_grid = {
'n_neighbors': [3, 5, 7, 10, 15],
'leaf_size': [3, 5, 7, 10, 15],
'algorithm': ['kd_tree']
}
grid_cv_KNN = GridSearchCV(estimator=KNeighborsClassifier(),
cv=cv,
scoring=scor,
param_grid=KNN_grid)
# grid_cv_KNN = RandomizedSearchCV(estimator=KNeighborsClassifier(), cv=cv,scoring=scor,param_distributions=KNN_grid)
grid_cv_KNN.fit(X.as_matrix(), y.as_matrix().ravel())
# print('grid_cv_KNN.best_params_: ',grid_cv_KNN.best_params_)
n_neighbors = grid_cv_KNN.best_params_['n_neighbors']
leaf_size = grid_cv_KNN.best_params_['leaf_size']
tuned_params = {
'n_neighbors': n_neighbors,
'leaf_size': leaf_size,
'max_window_size': knn_max_w_size
}
tuned_learn = KNN()
tuned_learn.set_params(**tuned_params)
tuned_learn.fit(X.as_matrix(), y.as_matrix().ravel())
elif cl_name == 'HoeffdingTree':
grace_period_range = np.array([25, 75, 150, 300])
tie_threshold_range = np.linspace(0.001, 1.0, 5)
split_confidence_range = np.linspace(0.000000001, 0.1, 5)
split_criterion_range = ['gini', 'info_gain', 'hellinger']
leaf_prediction_range = ['mc', 'nb', 'nba']
HT_grid = {
'grace_period': grace_period_range,
'tie_threshold': tie_threshold_range,
'split_confidence': split_confidence_range,
'split_criterion': split_criterion_range,
'leaf_prediction': leaf_prediction_range
}
grid_cv_HT = GridSearchCV(estimator=learn,
scoring=scor,
cv=cv,
param_grid=HT_grid)
# grid_cv_HT=RandomizedSearchCV(estimator=learn,scoring=scor,cv=cv,param_distributions=HT_grid)
grid_cv_HT.fit(X.as_matrix(), y.as_matrix().ravel())
# print('grid_cv_HT.best_params_: ',grid_cv_HT.best_params_)
tuned_params = grid_cv_HT.best_params_
tuned_learn = grid_cv_HT.best_estimator_
elif cl_name == 'NaiveBayes':
tuned_params = {'nominal_attributes': None}
tuned_learn = NaiveBayes()
tuned_learn.set_params(**tuned_params)
tuned_learn.fit(X.as_matrix(), y.as_matrix().ravel())
# print('Final tuned algorithm: ',tuned_learn)
return tuned_learn, tuned_params