def m007(label, cut_bound=10, drop_bound=1, max_level=10):
""" """
training = []
test = []
features = Counter()
stderr.write('constructing training set... ')
training_size = 0
for id, lab, history in iterdata(label):
training.append((id, lab, []))
for url in history.iterkeys():
if url[-1] == '/':
url = url[:-1]
training[training_size][2].append((url, 1))
features[url] += 1
training_size += 1
stderr.write('done\n')
stderr.write('modifying training set... \n')
for level in range(max_level, -1, -1):
stderr.write('\titeration with prefix length ' + str(level) + '... ')
for i in range(training_size):
for j in range(len(training[i][2])):
url = training[i][2][j][0]
if features[url] < cut_bound:
url = url_prefix(url, level)
training[i][2][j] = (url, 1)
features[url] += 1
stderr.write('done\n')
stderr.write('... done\n')
stderr.write('selecting most visited URLs as features... ')
features = set([f for f, c in features.iteritems() if c >= drop_bound])
features_map = dict([(f, c) for c, f in enumerate(features)])
stderr.write('done\n')
stderr.write('constructing test set... ')
test_size = 0
for id, lab, history in iterdata(label, 'test'):
test.append((id, lab, []))
for url in history.iterkeys():
if url[-1] == '/':
url = url[:-1]
for url in url_prefixes(url):
if url in features:
break
if url in features:
test[test_size][2].append((url, 1))
test_size += 1
stderr.write('done\n')
stderr.write('converting datasets to sparse matrices... ')
X, y = to_matrix(training, features_map)
X_test, y_test = to_matrix(test, features_map)
stderr.write('done\n')
return X, y, X_test, y_test, features_map
def m007(label, cut_bound=10, drop_bound=1, max_level=10):
""" """
training = []
test = []
features = Counter()
stderr.write('constructing training set... ')
training_size = 0
for id, lab, history in iterdata(label):
training.append((id, lab, []))
for url in history.iterkeys():
if url[-1] == '/':
url = url[:-1]
training[training_size][2].append((url, 1))
features[url] += 1
training_size += 1
stderr.write('done\n')
stderr.write('modifying training set... \n')
for level in range(max_level, -1, -1):
stderr.write('\titeration with prefix length ' + str(level) + '... ')
for i in range(training_size):
for j in range(len(training[i][2])):
url = training[i][2][j][0]
if features[url] < cut_bound:
url = url_prefix(url, level)
training[i][2][j] = (url, 1)
features[url] += 1
stderr.write('done\n')
stderr.write('... done\n')
stderr.write('selecting most visited URLs as features... ')
features = set([f for f, c in features.iteritems() if c >= drop_bound])
features_map = dict([(f, c) for c, f in enumerate(features)])
stderr.write('done\n')
stderr.write('constructing test set... ')
test_size = 0
for id, lab, history in iterdata(label, 'test'):
test.append((id, lab, []))
for url in history.iterkeys():
if url[-1] == '/':
url = url[:-1]
for url in url_prefixes(url):
if url in features:
break
if url in features:
test[test_size][2].append((url, 1))
test_size += 1
stderr.write('done\n')
stderr.write('converting datasets to sparse matrices... ')
X, y = to_matrix(training, features_map)
X_test, y_test = to_matrix(test, features_map)
stderr.write('done\n')
return X, y, X_test, y_test, features_map
def m008(label, lower_pop_bound=5):
""" """
training = []
test = []
features = Counter()
# Constructing training set.
# If http(s)://<X>/<folder> is visited, we say that
# http(s)://<X> is also visited, and apply this recursively.
stderr.write('constructing training set... ')
training_size = 0
for id, lab, history in iterdata(label):
training.append((id, lab, set()))
for url in history.iterkeys():
for x in url_prefixes(url):
training[training_size][2].add((x, 1))
features[x] += 1
training_size += 1
stderr.write('done\n')
stderr.write('selecting most visited URLs as features... ')
features = set([f for f, c in features.iteritems()
if c >= lower_pop_bound])
features_map = dict([(f, c) for c, f in enumerate(features)])
stderr.write('done\n')
stderr.write('constructing test set... ')
test_size = 0
for id, lab, history in iterdata(label, 'test'):
test.append((id, lab, set()))
for url in history.iterkeys():
for x in url_prefixes(url):
if x in features:
test[test_size][2].add((x, 1))
test_size += 1
stderr.write('done\n')
stderr.write('converting datasets to sparse matrices... ')
X, y = to_matrix(training, features_map)
X_test, y_test = to_matrix(test, features_map)
stderr.write('done\n')
return X, y, X_test, y_test, features_map
