def test_allPredictionsWrong(self):
# Prepare the SUT
# Create an artificial email classification dictionary
truth_dict = create_classification()
create_truth_and_prediction_file(truth_dict, invert_classes)
# Excercise the SUT
q = compute_quality_for_corpus(CORPUS_DIR)
# Assertions
self.assertEqual(q, 0.0)
def test_allPredictionsWrong(self):
# Prepare the SUT
# Create an artificial email classification dictionary
truth_dict = create_classification()
create_truth_and_prediction_file(truth_dict, invert_classes)
# Excercise the SUT
with replaced_open(): # Insist on explicit use of encoding
q = self.compute_quality_for_corpus(CORPUS_DIR)
# Assertions
self.assertEqual(q, 0.0)
def create_identical_truth_and_prediction_file():
"""
Create identical !truth.txt and !prediction.txt files in the corpus directory.
Here we assume that the corpus directory already exists.
"""
# Create an artificial email classification dictionary
class_dict = create_classification()
# Compile the filepaths
truth_filepath = os.path.join(CORPUS_DIR, TRUTH_FILENAME)
pred_filepath = os.path.join(CORPUS_DIR, PREDICTION_FILANAME)
# Save the same dictionary as both the !truth.txt and !prediction.txt
save_classification_to_file(class_dict, truth_filepath)
save_classification_to_file(class_dict, pred_filepath)
def test_allPredictionsHam_for10SpamsAnd10Hams(self):
# Prepare the SUT
# Create an artificial email classification dictionary
truth_dict = create_classification(n_items=20, n_spams=10)
create_truth_and_prediction_file(truth_dict, hams_only)
# Since there are 10 spams and 10 hams in the corpus,
# and predictions are all ham, then the confusion matrix
# shall have TN = 10 and FN = 10, zero positives.
# The modified accuracy then is:
expected_q = 10 / (10 + 10)
# Excercise the SUT
q = compute_quality_for_corpus(CORPUS_DIR)
# Assertions
self.assertEqual(q, expected_q)
def test_allPredictionsHam_for10SpamsAnd20Hams(self):
# Prepare the SUT
# Create an artificial email classification dictionary
truth_dict = create_classification(n_items=30, n_spams=10)
create_truth_and_prediction_file(truth_dict, hams_only)
# Since there are 10 spams and 20 hams in the corpus,
# and predictions are all ham, then the confusion matrix
# shall have TN = 20 and FN = 10, zero positives.
# The modified accuracy then is:
expected_q = 20 / (20 + 10)
# Excercise the SUT
with replaced_open(): # Insist on explicit use of encoding
q = self.compute_quality_for_corpus(CORPUS_DIR)
# Assertions
self.assertEqual(q, expected_q)
def create_inverse_truth_and_prediction_file():
"""
Create inverse !truth.txt and !prediction.txt files in the corpus directory.
Here we assume that the corpus directory already exists.
"""
# Create an artificial truth dictionary
truth_dict = create_classification()
# Create an inverted version of truth_dict
pred_dict = invert_classes(truth_dict)
# Compile the filepaths
truth_filepath = os.path.join(CORPUS_DIR, TRUTH_FILENAME)
pred_filepath = os.path.join(CORPUS_DIR, PREDICTION_FILANAME)
# Save the dictionaries in !truth.txt and !prediction.txt, respectively.
save_classification_to_file(truth_dict, truth_filepath)
save_classification_to_file(pred_dict, pred_filepath)
def test_1FP2FN_for10SpamsAnd20Hams(self):
# Prepare the SUT
# Create an artificial email classification dictionary
truth_dict = create_classification(n_items=30, n_spams=10)
create_truth_and_prediction_file(truth_dict, n_FP_n_FN)
# Since there are 10 spams and 20 hams in the corpus,
# and predictions are all correct except 1 FP and 2 FN,
# then the confusion matrix shall have
# FP = 1, FN = 2,
# TN = 19, TP = 8.
# The modified accuracy then is:
expected_q = (8 + 19) / (8 + 19 + 10 * 1 + 2)
# Excercise the SUT
with replaced_open(): # Insist on explicit use of encoding
q = self.compute_quality_for_corpus(CORPUS_DIR)
# Assertions
self.assertEqual(q, expected_q)
