def __init__(self):
Confusion.__init__(self)
self.symbolicExec = None
self.flaRecover = None
def tenFoldData(k, positive_training_size, negative_training_size, prior_positive_fold_probability, prior_negative_fold_probability):
"""Summary
Args:
k (TYPE): Fold number
positive_training_size (TYPE): The training size for the positive reviews in the fold method
negative_training_size (TYPE): The training size for the positive reviews in the fold method
prior_positive_fold_probability (TYPE): The probability of the positive reviews over both reviews for fold method
prior_negative_fold_probability (TYPE): The probability of the negative reviews over both reviews for fold method
Returns:
TYPE: Calculates the confusion_matrix and accuracy of the reviews using the 10 fold cross validation method
"""
total_tp = []
total_fp = []
total_tn = []
total_fn = []
total_accuracy = []
for f in xrange(0, k):
confusion_matrix = Confusion()
positive_training_list = glob.glob(
'review_polarity/txt_sentoken/pos/cv[!' + str(f) + ']*.txt')
new_positive_training_list = sentimentList(positive_training_list)
positive_sum_training_list = sentimentSumList(
new_positive_training_list)
positive_probability_training_list = sentimentProbabilityList(
positive_sum_training_list, positive_training_size)
positive_testing_list = glob.glob(
'review_polarity/txt_sentoken/pos/cv' + str(f) + '*.txt')
new_positive_testing_list = sentimentList(positive_testing_list)
negative_training_list = glob.glob(
'review_polarity/txt_sentoken/neg/cv[!' + str(f) + ']*.txt')
new_negative_training_list = sentimentList(negative_training_list)
negative_sum_training_list = sentimentSumList(
new_negative_training_list)
negative_probability_training_list = sentimentProbabilityList(
negative_sum_training_list, negative_training_size)
negative_testing_list = glob.glob(
'review_polarity/txt_sentoken/neg/cv' + str(f) + '*.txt')
new_negative_testing_list = sentimentList(negative_testing_list)
for positive_vector in new_positive_testing_list:
polarity = naiveBayesClassifier(positive_vector, prior_positive_fold_probability,
prior_negative_fold_probability, positive_probability_training_list, negative_probability_training_list)
if polarity:
confusion_matrix.incrementTP()
else:
confusion_matrix.incrementFP()
for negative_vector in new_negative_testing_list:
polarity = naiveBayesClassifier(negative_vector, prior_positive_fold_probability,
prior_negative_fold_probability, positive_probability_training_list, negative_probability_training_list)
if polarity:
confusion_matrix.incrementFN()
else:
confusion_matrix.incrementTN()
total_tp.append(confusion_matrix.getTP())
total_fp.append(confusion_matrix.getFP())
total_tn.append(confusion_matrix.getTN())
total_fn.append(confusion_matrix.getFN())
accuracy = (confusion_matrix.getTP() + confusion_matrix.getTN()) / float(confusion_matrix.getTP() +
confusion_matrix.getFP() + confusion_matrix.getTN() + confusion_matrix.getFN())
total_accuracy.append(accuracy)
print "True Positive: " + str(sum(total_tp))
print "False Positive: " + str(sum(total_fp))
print "True Negative: " + str(sum(total_tn))
print "False Negative: " + str(sum(total_fn))
print "10-Fold Cross Validation Method"
print "Accuracy: " + str(sum(total_accuracy) / k)
def completeData(positive_review_count, negative_review_count, new_positive_list, new_negative_list, prior_positive_probability, prior_negative_probability):
"""Summary
Args:
positive_review_count (TYPE): Total number of positive reviews
negative_review_count (TYPE): Total number of negative reviews
new_positive_list (TYPE): Binary list of lists for the positive reviews
new_negative_list (TYPE): Binary list of lists for the negative reviews
prior_positive_probability (TYPE): The probability of the positive reviews over both reviews
prior_negative_probability (TYPE): The probability of the negative reviews over both reviews
Returns:
TYPE: Runs the complete data set as both the training and testing sets to get the confusion matrix and accuracy
"""
negative_sum_list = sentimentSumList(new_negative_list)
negative_probability_list = sentimentProbabilityList(
negative_sum_list, negative_review_count)
positive_sum_list = sentimentSumList(new_positive_list)
positive_probability_list = sentimentProbabilityList(
positive_sum_list, positive_review_count)
print positive_probability_list
print negative_probability_list
confusion_matrix = Confusion()
for positive_vector in new_positive_list:
polarity = naiveBayesClassifier(positive_vector, prior_positive_probability,
prior_negative_probability, positive_probability_list, negative_probability_list)
if polarity:
confusion_matrix.incrementTP()
else:
confusion_matrix.incrementFP()
for negative_vector in new_negative_list:
polarity = naiveBayesClassifier(negative_vector, prior_positive_probability,
prior_negative_probability, positive_probability_list, negative_probability_list)
if polarity:
confusion_matrix.incrementFN()
else:
confusion_matrix.incrementTN()
accuracy = (confusion_matrix.getTP() + confusion_matrix.getTN()) / float(confusion_matrix.getTP() +
confusion_matrix.getFP() + confusion_matrix.getTN() + confusion_matrix.getFN())
print "True Positive: " + str(confusion_matrix.getTP())
print "False Positive: " + str(confusion_matrix.getFP())
print "True Negative: " + str(confusion_matrix.getTN())
print "False Negative: " + str(confusion_matrix.getFN())
print "Complete Data Method"
print "Accuracy: " + str(accuracy)
X_train, X_test, y_train, y_test = train_test_split(
synthetic[['x1', 'x2']],
synthetic[['y']],
test_size=test_size,
stratify=synthetic[['y']])
# scatter_plot(X_train, y_train, 'Train set')
# scatter_plot(X_test, y_test, 'Test set')
print('GaussianBayes:')
bc = GaussianBayesClassifier()
bc.fit(X_train, y_train)
pred = bc.evaluate(X_test)
real = list(y_test['y'])
confusion = Confusion.from_wrong_preds(['A1', 'A2'], pred, real, {
'A1': 50,
'A2': 50
})
print(confusion)
print('ParzenBayes:')
pc = ParzenBayesClassifier()
pc.fit(X_train, y_train)
pred = pc.evaluate(X_test)
real = list(y_test['y'])
confusion = Confusion.from_wrong_preds(['A1', 'A2'], pred, real, {
'A1': 50,
'A2': 50
})
print(confusion)
#####################################################################################