class FeaturesExtractor(object):
def __init__(self, user):
self.user = user
self.preprocessor = Preprocessor(self.user.tweets_text)
self.features = []
self.features.append(user.followers_count)
self.features.append(user.following_count)
self.features.append(user.total_tweets)
self.features.append(user.mentions_count)
self.features.append(user.replies_count)
self.features.append(user.hashtags_count)
self.features.append(user.links_count)
self.features.append(user.words_count)
self.features.append(user.tweets_count)
tokens = self.preprocessor.remove_stop_words_and_urls()
self.features.append(len(tokens))
sentiment = self.preprocessor.vader()
for k in sorted(sentiment):
self.features.append(sentiment[k])
mrc_location = os.getcwd() + '/resources/mrc/1054'
mrc = MrcService(mrc_location)
self.features.extend(mrc.get_vector(tokens))
def get_features(self):
return self.features
def predict(self, username):
prediction_dict = {}
solr_user = self.get_solr_user(username)
if solr_user:
print 'got data'
prediction_dict = solr_user
else:
user = self.ie.extract(username, self.tweet_count)
user_fe = FeaturesExtractor(user)
user_features = user_fe.get_features()
predicted = self.pp.predict(user_features)
preprocessor = Preprocessor(user.tweets_text)
prediction_dict = {
username: {
'f': user_features,
'o': predicted['o'],
'c': predicted['c'],
'e': predicted['e'],
'a': predicted['a'],
'n': predicted['n'],
'tweets': user.tweets_text,
'top_words': preprocessor.most_used_words(max_count=self.search_config['top_words']),
'hashtags': preprocessor.most_used_hashtags(max_count=self.search_config['top_hashtags']),
'bigrams': preprocessor.most_used_bigrams(max_count=self.search_config['top_bigrams'])
}
}
if prediction_dict[username]['hashtags'] == '':
prediction_dict[username]['hashtags'] = '$'
if prediction_dict[username]['bigrams'] == '':
prediction_dict[username]['bigrams'] = '$'
if prediction_dict[username]['top_words'] == '':
prediction_dict[username]['top_words'] = '$'
if prediction_dict[username]['tweets'] == '':
prediction_dict[username]['tweets'] = '$'
solr_dict = FeaturesConverter.convert_features_to_solr(prediction_dict)
self.solr.addUser(solr_dict[0])
similar = self.solr.getSimilarUsers(username, tf=self.search_config['tf'], df=self.search_config['df'], count=self.search_config['similar'])
prediction_dict[username].update({
'similar': similar,
'top_words': self.split_data(prediction_dict[username]['top_words']),
'hashtags': self.split_data(prediction_dict[username]['hashtags']),
'bigrams': self.split_bigrams(prediction_dict[username]['bigrams']),
})
return prediction_dict
def load_users(self):
i = 0
to_remove = []
for username in self.data:
if i == self.train_count:
break
solr_data = self.get_solr_user(username)
if solr_data:
self.data[username] = solr_data[username]
i += 1
continue
try:
user = self.ie.extract(username, self.tweet_count)
fe = FeaturesExtractor(user)
self.data[username]['f'] = fe.get_features()
if user.tweets_text == '':
continue
self.data[username]['tweets'] = user.tweets_text
preprocessor = Preprocessor(user.tweets_text)
self.data[username]['top_words'] = preprocessor.most_used_words(max_count=self.search_config['top_words'])
self.data[username]['hashtags'] = preprocessor.most_used_hashtags(max_count=self.search_config['top_hashtags'])
self.data[username]['bigrams'] = preprocessor.most_used_bigrams(max_count=self.search_config['top_bigrams'])
if self.data[username]['hashtags'] == '':
self.data[username]['hashtags'] = '$'
if self.data[username]['bigrams'] == '':
self.data[username]['bigrams'] = '$'
if self.data[username]['top_words'] == '':
self.data[username]['top_words'] = '$'
solr_dict = FeaturesConverter.convert_features_to_solr({username: self.data[username]})
self.solr.addUser(solr_dict[0])
i += 1
except ZeroDivisionError:
if username in self.data:
to_remove.append(username)
except TweepError:
if username in self.data:
to_remove.append(username)
for uname in to_remove:
if uname in self.data:
del self.data[uname]
