abduakhatov / Tweet-Classifier

Tweet crawling and classification.

Tweet-Classifier

---

Python Library to fetch tweets from any twitter handler and classify those fetched tweets into 6 different topics.

Dependencies:

• nltk
• tweepy
• scikit-learn

Usage:

First add your twitter api keys in the twc/Tweetifier.py file.

from twc import Tweetifier

#initialize the object
T = Tweetifier("paraazz", 100)

#crawl or fetch the tweets max: 3200
T.crawl()
crawled_tweets = T.tweets

#classify the tweets topic wise 
#(Multinomial Naive Bayes Classifier accuracy: 78%)
T.classify()
tweets_topic = T.topic_bucket

Training Dataset and Model:

Dataset was created by fetching titles of different subreddit relating to 6 main following categories.

• technology
• business
• politics
• entertainment
• sports
• health

---

To refresh the dataset with new headlines, run the script in dir twc/data/:

$ python3 fetch_data.py

and train the model again in twc/

$ python3 model_train.py

The classifier being used here is Multinomial Naive Bayes Classifier with accuracy of 78%.

Other classifiers Accuracy (On this dataset):

• Naive Bayes Classifier: 72%
• SVC: 74%