def test_predict_prediction_and_score(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
prediction, score = model.predict(self.a_very_close_game)
# Correct Prediction Output
self.assertEqual(prediction, 0)
# Correct Score Output
np.testing.assert_array_almost_equal(
score[0], self.correct_a_very_close_game_score)
def test_train_model_params(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
# Correct Model
np.testing.assert_array_almost_equal(model.priors, self.correct_priors)
np.testing.assert_array_almost_equal(model.label_counts,
self.correct_label_count)
np.testing.assert_array_almost_equal(model.empty_likelihoods,
self.correct_empty_likelihoods)
np.testing.assert_array_almost_equal(model.likelihoods,
self.correct_likelihoods)
def test_update_dictionary_shorten_dictionary(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
model.update_dictionary(self.sports_map, self.sports_map_shortened)
prediction, score = model.predict(self.a_very_close_game_short)
# Correct Model Parameter Updates
np.testing.assert_array_almost_equal(model.priors, self.correct_priors)
np.testing.assert_array_almost_equal(model.label_counts,
self.correct_label_count)
np.testing.assert_array_almost_equal(
model.empty_likelihoods, self.correct_shortened_empty_likelihoods)
np.testing.assert_array_almost_equal(
model.likelihoods, self.correct_shortened_likelihoods)
# Correct Scores
np.testing.assert_array_almost_equal(
score[0], self.correct_shortened_a_very_close_game_score)
# Correct Prediction
self.assertEqual(prediction[0], 0)
def test_update_add_more_training_data(self):
model = Multinomial()
model.train(self.sports_labels[0:4], self.sports_data[0:4])
model.update(np.array([self.sports_labels[4]]),
np.array([self.sports_data[4]]))
prediction, score = model.predict(self.a_very_close_game)
# Correct Model Parameter Updates
np.testing.assert_array_almost_equal(model.priors, self.correct_priors)
np.testing.assert_array_almost_equal(model.label_counts,
self.correct_label_count)
np.testing.assert_array_almost_equal(model.empty_likelihoods,
self.correct_empty_likelihoods)
np.testing.assert_array_almost_equal(model.likelihoods,
self.correct_likelihoods)
# Correct New Scores
np.testing.assert_array_almost_equal(
score[0], self.correct_a_very_close_game_score)
# Correct Prediction
self.assertEqual(prediction[0], 0)
def test_update_dictionary_new_map_not_dict(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.assertRaises(TypeError, model.update_dictionary, self.sports_map,
list(self.sports_map_shortened))
def test_update_dictionary_old_map_not_same_size_as_training_data(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.sports_map["test"] = 15
self.assertRaises(ValueError, model.update_dictionary, self.sports_map,
self.sports_map_shortened)
def test_update_dictionary_old_map_wrong_keys(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.sports_map[0] = 1
self.assertRaises(TypeError, model.update_dictionary, self.sports_map,
self.sports_map_shortened)
def test_update_called_before_training(self):
model = Multinomial()
self.assertRaises(ValueError, model.update, self.sports_labels,
self.sports_data)
def test_update_new_train_data_not_2D(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.assertRaises(ValueError, model.update, self.sports_labels,
self.sports_data[:, np.newaxis])
def test__update_new_train_data_not_same_dim_as_training_data(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.assertRaises(ValueError, model.update, self.sports_labels,
self.sports_data[:, 0:2])
def test_update_dictionary_called_before_training(self):
model = Multinomial()
self.assertRaises(ValueError, model.update_dictionary, self.sports_map,
self.sports_map_extra)
def test_train_labels_not_1D(self):
model = Multinomial()
self.assertRaises(ValueError, model.train,
self.sports_labels[:, np.newaxis], self.sports_data)
def test_predict_called_before_training(self):
model = Multinomial()
self.assertRaises(ValueError, model.predict, self.a_very_close_game)
def test_update_dictionary_new_map_wrong_values(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.sports_map_shortened["test"] = "fun"
self.assertRaises(TypeError, model.update_dictionary, self.sports_map,
self.sports_map_shortened)
def test_predict_test_data_not_same_dim_as_training_data(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.assertRaises(ValueError, model.predict,
self.a_very_close_game[:, 0:2])
def test_predict_test_data_not_2D(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.assertRaises(ValueError, model.predict,
np.array(self.a_very_close_game[0]))
def test_predict_test_data_not_numpy_array(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.assertRaises(TypeError, model.predict,
list(self.a_very_close_game))
def test_train_training_data_not_numpy_array(self):
model = Multinomial()
self.assertRaises(TypeError, model.train, self.sports_labels,
list(self.sports_data))
def test_train_training_data_not_2D(self):
model = Multinomial()
self.assertRaises(ValueError, model.train, self.sports_labels,
self.sports_data[:, np.newaxis])
def test_update_dictionary_new_map_is_empty(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.assertRaises(ValueError, model.update_dictionary, self.sports_map,
{})
def test_update_number_of_labels_and_docs_differ(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.assertRaises(ValueError, model.update, self.sports_labels[0:4],
self.sports_data)
def test_update_new_train_data_not_numpy_array(self):
model = Multinomial()
model.train(self.sports_labels, self.sports_data)
self.assertRaises(TypeError, model.update, self.sports_labels,
list(self.sports_data))
matches += 1
print("- Dictionary")
print("Accuracy: " + str(matches / len(labels)))
################
# Vector Model #
################
wine_vector_data = wine_data["vectors"]
index_map = wine_vector_data["word_map"]
reviews_as_vector = wine_vector_data["vectors"]
labels_as_vectors = wine_vector_data["labels"]
model_vect = VectMultinomial()
model_vect.train(labels_as_vectors, reviews_as_vector)
vect_predictions, vect_scores = model_vect.predict(reviews_as_vector)
accuracy = np.sum(
vect_predictions == labels_as_vectors) / labels_as_vectors.shape[0]
print("- Vector")
print("Accuracy: " + str(accuracy))
#####################
# Spam Data Example #
#####################
print("\nSpam Data Example")
with open('./sample_data/kaggle/spam/sms_spam.pkl', 'rb') as f:
