def main(filename):
# initialize global data
labels, _, comments = util.get_comments_data(filename)
vec = TfidfVectorizer(
ngram_range = (1, 2),
strip_accents = None,
charset_error = 'ignore',
stop_words = None,
min_df = 2
)
labels, instances = numpy.array(labels), vec.fit_transform(comments)
random.seed(0)
# cross validate
n_models = 9; cv = 5; cv_accuracy = list();
for i in xrange(0, cv):
print "Iteration #" + str(i) + "..."
# initialize training/testing data
cv_data = cross_validation.train_test_split(instances, labels, test_size = 0.3, random_state = i)
x_training = cv_data[0]
x_testing = cv_data[1]
y_training = cv_data[2]
y_testing = cv_data[3]
# initialize the classifier
clf = BaggingSVM(n_models)
clf.fit(x_training, y_training)
# measure prediction accuracy
cv_accuracy.append(clf.score(x_testing, y_testing))
print "Scores => " + str(cv_accuracy)
print "Mean => " + str(numpy.mean(cv_accuracy))
def main(infile, outfile):
"""read from infile, write to outfile"""
print "Reading from %s..." % infile
labels, timestamps, comments = get_comments_data(infile)
print "Parsing..."
vec = TfidfVectorizer(ngram_range=(1, 2),
strip_accents=None,
charset_error="ignore",
stop_words=None,
min_df=2)
vec.fit(comments)
print "Processing and writing to %s..." % outfile
f = open(outfile, 'w')
counter = 0
comments = vec.transform(comments)
rows, cols = comments.get_shape()
for row in xrange(0, rows):
if counter % 100 == 0:
print counter
buf, indices = list(), comments[row].indices.tolist()
indices.sort()
buf.append(str(float(labels[row])))
for col in indices:
buf.append("%d:%.3f" % (col, comments[row, col]))
buf.append("\n")
f.write(" ".join(buf))
counter = counter + 1
f.close()
print "Done!"
def main(infile, outfile):
"""read from infile, write to outfile"""
print "Reading from %s..." % infile
labels, timestamps, comments = get_comments_data(infile)
print "Parsing..."
vec = TfidfVectorizer(
ngram_range = (1, 2),
strip_accents = None,
charset_error = "ignore",
stop_words = None,
min_df = 2
)
vec.fit(comments)
print "Processing and writing to %s..." % outfile
f = open(outfile, 'w')
counter = 0
comments = vec.transform(comments)
rows, cols = comments.get_shape()
for row in xrange(0, rows):
if counter % 100 == 0:
print counter
buf, indices = list(), comments[row].indices.tolist()
indices.sort()
buf.append(str(float(labels[row])))
for col in indices:
buf.append("%d:%.3f" % (col, comments[row, col]))
buf.append("\n")
f.write(" ".join(buf))
counter = counter + 1
f.close()
print "Done!"
