def test_logistic_regression_model_train_evaluate():
data, labels = load_iris(return_X_y=True)
randomize = np.arange(len(labels))
np.random.shuffle(randomize)
data = data[randomize, ]
labels = labels[randomize]
dim = 100
train_data = data[0:dim, ]
train_labels = labels[0:dim]
test_data = data[dim:, ]
test_labels = labels[dim:]
model = LogisticRegression(max_iter=150)
lgr = LinearClassifierModel(n_features=np.shape(train_data)[1],
classes=np.unique(train_labels),
model=model)
lgr.train(data=train_data, labels=train_labels)
evaluation = np.array(lgr.evaluate(data=test_data, labels=test_labels))
performance = lgr.performance(data=test_data, labels=test_labels)
prediction = lgr.predict(data=test_data)
model_params = lgr.get_model_params()
lgr_ref = LogisticRegression(max_iter=150).fit(train_data, train_labels)
prediction_ref = lgr_ref.predict(test_data)
assert np.array_equal(model_params[0], lgr_ref.intercept_)
assert np.array_equal(model_params[1], lgr_ref.coef_)
assert np.array_equal(prediction, prediction_ref)
assert np.array_equal(evaluation, np.array((metrics.balanced_accuracy_score(test_labels, prediction_ref),\
metrics.cohen_kappa_score(test_labels, prediction_ref))))
assert performance == metrics.balanced_accuracy_score(
test_labels, prediction_ref)
def test_linearSVC_model_train_evaluate():
data, labels = load_iris(return_X_y=True)
randomize = np.arange(len(labels))
np.random.shuffle(randomize)
data = data[randomize, ]
labels = labels[randomize]
dim = 100
train_data = data[0:dim, ]
train_labels = labels[0:dim]
test_data = data[dim:, ]
test_labels = labels[dim:]
model = LinearSVC(random_state=123)
svc = LinearClassifierModel(n_features=np.shape(train_data)[1],
classes=np.unique(train_labels),
model=model)
svc.train(data=train_data, labels=train_labels)
evaluation = np.array(svc.evaluate(data=test_data, labels=test_labels))
performance = svc.performance(data=test_data, labels=test_labels)
prediction = svc.predict(data=test_data)
model_params = svc.get_model_params()
svc_ref = LinearSVC(random_state=123).fit(train_data, train_labels)
prediction_ref = svc_ref.predict(test_data)
assert np.array_equal(model_params[0], svc_ref.intercept_)
assert np.array_equal(model_params[1], svc_ref.coef_)
assert np.array_equal(prediction, prediction_ref)
assert np.array_equal(
evaluation,
np.array((metrics.balanced_accuracy_score(test_labels, prediction_ref),
metrics.cohen_kappa_score(test_labels, prediction_ref))))
assert performance == metrics.balanced_accuracy_score(
test_labels, prediction_ref)