def test_old_style_commands(self):
my_learner = baseline_learner.BaselineLearner
session1 = manager.ToolkitSession(arff=self.arff_path, learner=my_learner,
eval_method="random", eval_parameter=.7, label_count=1)
session2 = manager.ToolkitSession(arff=self.arff_path, learner=my_learner,
eval_method="static", eval_parameter=self.arff_path, label_count=1)
session3 = manager.ToolkitSession(arff=self.arff_path, learner=my_learner,
eval_method="cross", eval_parameter=10, label_count=1)
def test_confusion_matrix(self):
my_learner = baseline_learner.BaselineLearner
data_arff = arff.Arff(self.arff_path, label_count=1)
session = manager.ToolkitSession(arff=self.arff_path, learner=my_learner, data=data_arff, example_hyperparameter=.5, label_count=1)
session.train()
cm = session.learner.get_confusion_matrix(data_arff.get_features(), data_arff.get_labels())
self.assertEqual('383',cm[-1,-1])
def test_train(self):
my_learner = baseline_learner.BaselineLearner
session = manager.ToolkitSession(arff=self.arff_path, learner=my_learner, eval_method=None,
eval_parameter=None, print_confusion_matrix=False, normalize=False,
random_seed=None, label_count=1)
session.train()
self.assertAlmostEqual(session.training_accuracy[0],0.55507246)
def test_train_test_manual(self):
my_learner = baseline_learner.BaselineLearner
session = manager.ToolkitSession(arff=self.arff_path, learner=my_learner, eval_method=None,
eval_parameter=None, print_confusion_matrix=True, normalize=False,
random_seed=None, label_count=1)
train_features, train_labels, test_features, test_labels = session.training_test_split(.7) # 70% training
session.train(train_features, train_labels)
session.test(test_features, test_labels)
def test_learner_kwargs(self):
"""
Returns:
"""
my_learner = baseline_learner.BaselineLearner
data_arff = arff.Arff(self.arff_path, label_count=1)
session = manager.ToolkitSession(arff=self.arff_path, learner=my_learner, data=data_arff, example_hyperparameter=.5, label_count=1)
self.assertEqual(session.learner.data_shape,(690,16))
self.assertEqual(session.learner.example_hyperparameter, .5)
def test_handle_numpy_array(self):
""" If arrays are passed to measure accuracy instead of arff objects
"""
my_learner = baseline_learner.BaselineLearner
session = manager.ToolkitSession(arff=self.arff_path, learner=my_learner, label_count=1)
train_features, train_labels, test_features, test_labels = session.training_test_split(.7) # 70% training
session.train(train_features, train_labels)
session.learner.measure_accuracy(test_features.data, test_labels.data) # will test accuracy, assumes no nominal variables
session.learner.measure_accuracy(test_features.data, test_labels.data, eval_method="sse")
# Test naive confusion matrix
session.learner.measure_accuracy(test_features.data, test_labels.data)
# Print as arff
print(features)
# Print Numpy array
print(features.data)
# Print first row as Numpy array
print(features[0, :])
# Get all labels as numpy array using slicing
labels = credit_approval.get_labels()[:]
# Manual Training/Test
my_learner = baseline_learner.BaselineLearner
session = manager.ToolkitSession(arff=credit_approval, learner=my_learner)
train_features, train_labels, test_features, test_labels = session.training_test_split(
.7) # 70% training
session.train(train_features, train_labels)
session.test(test_features, test_labels)
print(session.training_accuracy)
# Pass on hyperparameters to learner
session = manager.ToolkitSession(arff=credit_approval,
learner=my_learner,
data=credit_approval,
example_hyperparameter=.5)
print(session.learner.example_hyperparameter) # .5
# Automatic
session2 = manager.ToolkitSession(arff=credit_approval,
def test_explicit_construction(self):
data_arff = arff.Arff(self.arff_path, label_count=1)
my_learner = baseline_learner.BaselineLearner(data_arff)
session = manager.ToolkitSession(arff=self.arff_path, learner=my_learner, eval_method="training",
eval_parameter=None, print_confusion_matrix=False, normalize=False,
random_seed=None, label_count=1)
def test_sse(self):
my_learner = baseline_learner.BaselineLearner
session = manager.ToolkitSession(arff=self.arff_path, learner=my_learner, label_count=1)
train_features, train_labels, test_features, test_labels = session.training_test_split(.7) # 70% training
session.train(train_features, train_labels)
session.learner.measure_accuracy(test_features, test_labels, eval_method="sse")