def test_fit_defaultclassifier(data):
clf = CTClassifier(random_state=10)
clf.fit(data['random_data'], data['random_labels'])
y_pred_test = clf.predict(data['random_test'])
y_pred_prob = clf.predict_proba(data['random_test'])
truth = [2., 3., 2., 2., 2.]
truth_proba = [[0.88555037, 0.11444963], [0.05650123, 0.94349877],
[0.50057741, 0.49942259], [0.89236186, 0.10763814],
[0.95357416, 0.04642584]]
for i in range(data['N_test']):
assert y_pred_test[i] == truth[i]
for i in range(data['N_test']):
for j in range(2):
assert abs(y_pred_prob[i, j] - truth_proba[i][j]) < 0.00000001
def test_predict_num_iter(data):
random_seed = 10
gnb1 = GaussianNB()
gnb2 = GaussianNB()
clf = CTClassifier(gnb1, gnb2, num_iter=9, random_state=random_seed)
clf.fit(data['random_data'], data['random_labels'])
y_pred_test = clf.predict(data['random_test'])
y_pred_prob = clf.predict_proba(data['random_test'])
truth = [2., 3., 2., 2., 2.]
truth_proba = [[0.88555037, 0.11444963], [0.05650123, 0.94349877],
[0.50057741, 0.49942259], [0.89236186, 0.10763814],
[0.95357416, 0.04642584]]
for i in range(data['N_test']):
assert y_pred_test[i] == truth[i]
for i in range(data['N_test']):
for j in range(2):
assert abs(y_pred_prob[i, j] - truth_proba[i][j]) < 0.00000001
def test_predict_set_n(data):
random_seed = 10
gnb1 = GaussianNB()
gnb2 = GaussianNB()
clf = CTClassifier(gnb1, gnb2, n=9, random_state=random_seed)
clf.fit(data['random_data'], data['random_labels'])
y_pred_test = clf.predict(data['random_test'])
y_pred_prob = clf.predict_proba(data['random_test'])
truth = [3., 3., 2., 3., 3.]
truth_proba = [[0.29020704, 0.70979296], [0.44024614, 0.55975386],
[0.5710383, 0.4289617], [0.37366059, 0.62633941],
[0.22157484, 0.77842516]]
for i in range(data['N_test']):
assert y_pred_test[i] == truth[i]
for i in range(data['N_test']):
for j in range(2):
assert abs(y_pred_prob[i, j] - truth_proba[i][j]) < 0.00000001
def test_predict_set_p(data):
random_seed = 10
gnb1 = GaussianNB()
gnb2 = GaussianNB()
clf = CTClassifier(gnb1, gnb2, p=12, random_state=random_seed)
clf.fit(data['random_data'], data['random_labels'])
y_pred_test = clf.predict(data['random_test'])
y_pred_prob = clf.predict_proba(data['random_test'])
truth = [3., 3., 3., 3., 3.]
truth_proba = [[0.31422418, 0.68577582], [0.40938282, 0.59061718],
[0.48448605, 0.51551395], [0.38853225, 0.61146775],
[0.22972488, 0.77027512]]
for i in range(data['N_test']):
assert y_pred_test[i] == truth[i]
for i in range(data['N_test']):
for j in range(2):
assert abs(y_pred_prob[i, j] - truth_proba[i][j]) < 0.00000001
def test_predict_check_p_n(data):
labels1 = np.zeros(100, )
labels1[:5] = 4 # 5 "negative"
labels1[5:15] = 6 # 10 "positive"
labels1[15:] = np.nan
clf = CTClassifier()
clf.fit(data['random_data'], labels1)
assert clf.p_ == 2
assert clf.n_ == 1
labels2 = np.zeros(100, )
labels2[:5] = 6 # 5 "positive"
labels2[5:15] = 4 # 10 "negative"
labels2[15:] = np.nan
clf = CTClassifier()
clf.fit(data['random_data'], labels2)
assert clf.p_ == 1
assert clf.n_ == 2
labels1 = np.zeros(100, )
labels1[:5] = 4 # 5 "negative"
labels1[5:15] = 6 # 10 "positive"
labels1[15:] = np.nan
clf = CTClassifier(p=4, n=3)
clf.fit(data['random_data'], labels1)
assert clf.p_ == 4
assert clf.n_ == 3
def test_predict_unlabeled_pool_size(data):
random_seed = 10
gnb1 = GaussianNB()
gnb2 = GaussianNB()
clf = CTClassifier(gnb1,
gnb2,
unlabeled_pool_size=20,
random_state=random_seed)
clf.fit(data['random_data'], data['random_labels'])
y_pred_test = clf.predict(data['random_test'])
y_pred_prob = clf.predict_proba(data['random_test'])
truth = [2., 3., 2., 2., 2.]
truth_proba = [[0.55708013, 0.44291987], [0.29591617, 0.70408383],
[0.50441055, 0.49558945], [0.99276393, 0.00723607],
[0.95221514, 0.04778486]]
for i in range(data['N_test']):
assert y_pred_test[i] == truth[i]
for i in range(data['N_test']):
for j in range(2):
assert abs(y_pred_prob[i, j] - truth_proba[i][j]) < 0.00000001