def main():
### Read arguments
if len(sys.argv) != 4:
print(USAGE % sys.argv[0])
testing_folder = sys.argv[1]
(spam_folder, ham_folder) = sys.argv[2:4]
### Learn the distributions
print("Training...")
file_lists = []
for folder in (spam_folder, ham_folder):
file_lists.append(util.get_files_in_folder(folder))
(log_probabilities_by_category, log_priors_by_category) = \
learn_distributions(file_lists)
#print(log_probabilities_by_category)
# Here, columns and rows are indexed by 0 = 'spam' and 1 = 'ham'
# rows correspond to true label, columns correspond to guessed label
performance_measures = np.zeros([2,2])
### Classify and measure performance
print("Testing...")
idx = 1
for filename in (util.get_files_in_folder(testing_folder)):
print(idx)
print(filename)
idx += 1
## Classify
label = classify_email(filename,
log_probabilities_by_category,
log_priors_by_category)
## Measure performance
# Use the filename to determine the true label
base = os.path.basename(filename)
true_index = ('ham' in base)
guessed_index = (label == 'ham')
performance_measures[true_index, guessed_index] += 1
# Uncomment this line to see which files your classifier
# gets right/wrong:
#print("%s : %s" %(label, filename))
template="You correctly classified %d out of %d spam emails, and %d out of %d ham emails."
# Correct counts are on the diagonal
correct = np.diag(performance_measures)
# totals are obtained by summing across guessed labels
totals = np.sum(performance_measures, 1)
print(template % (correct[0],
totals[0],
correct[1],
totals[1]))
def SplitMultipageTiff(directory):
for i in get_files_in_folder(directory):
if i.endswith('.tif'):
outfile = i.replace('.tif', '-%d.tif')
args = [
'convert',
i,
outfile
]
subprocess.call(args, shell=True)
plt.ylabel('Logistic Regression Cost')
plt.show()
return theta
if __name__ == '__main__':
### Read arguments
if len(sys.argv) != 4:
print(USAGE % sys.argv[0])
testing_folder = sys.argv[1]
(spam_folder, ham_folder) = sys.argv[2:4]
file_lists = []
for folder in (spam_folder, ham_folder):
if NUM_EXAMPLES > 0:
file_lists.append(util.get_files_in_folder(folder)[:NUM_EXAMPLES])
else:
file_lists.append(util.get_files_in_folder(folder))
print("Extracting Features and Training...")
theta, all_words = train_logistic(file_lists)
# # Here, columns and rows are indexed by 0 = 'spam' and 1 = 'ham'
# # rows correspond to true label, columns correspond to guessed label
performance_measures = np.zeros([2, 2])
### Classify and measure performance
print("Testing...")
idx = 1
for filename in (util.get_files_in_folder(testing_folder)):
idx += 1
## Classify
posterior = np.array(log_likelihoods) + np.array(log_prior_by_category)
winner = np.argmax(posterior)
return names[winner]
if __name__ == '__main__':
### Read arguments
if len(sys.argv) != 4:
print USAGE % sys.argv[0]
testing_folder = sys.argv[1]
(spam_folder, ham_folder) = sys.argv[2:4]
### Learn the distributions
print("Training...")
file_lists = []
for folder in (spam_folder, ham_folder):
file_lists.append(util.get_files_in_folder(folder))
(log_probabilities_by_category, log_priors_by_category) = \
learn_distributions(file_lists)
# Here, columns and rows are indexed by 0 = 'spam' and 1 = 'ham'
# rows correspond to true label, columns correspond to guessed label
performance_measures = np.zeros([2,2])
### Classify and measure performance
print("Testing...")
for filename in (util.get_files_in_folder(testing_folder)):
## Classify
label = classify_message(filename,
log_probabilities_by_category,
log_priors_by_category,
['spam', 'ham'])
sum_log_prob_given_spam += math.log(p_d.get(word, 0.5))
sum_log_prob_given_ham += math.log(q_d.get(word, 0.5))
result = ("spam", "ham")[sum_log_prob_given_spam/sum_log_prob_given_ham > 0.99]
return (result, [sum_log_prob_given_spam, sum_log_prob_given_ham])
if __name__ == '__main__':
spam_folder = "data/spam"
ham_folder = "data/ham"
test_folder = "data/testing"
# Get training data
file_lists = []
for folder in (spam_folder, ham_folder):
file_lists.append(util.get_files_in_folder(folder))
# Learn the distributions
probabilities_by_category = learn_distributions(file_lists)
# prior class distribution
priors_by_category = [0.5, 0.5]
# Store the classification results
performance_measures = np.zeros([2,2])
# Explanation of performance_measures:
# columns and rows are indexed by 0 = 'spam' and 1 = 'ham'
# rows correspond to true label, columns correspond to guessed label
# to be more clear, performance_measures = [[p1 p2]
# [p3 p4]]
import sys
import subprocess
from util import get_files_in_folder
directory = sys.argv[1]
for i in get_files_in_folder(directory):
if i.endswith('.tif'):
outfile = i.replace('.tif', '-%d.tif')
args = ['convert', i, outfile]
subprocess.call(args, shell=True)
