Twitter Trust Propagation

Makes predictions of trustworthiness of people in your network based on who their connections are. Downloads timelines and connections, and saves each user's data to a separate json file. Assigns a score to each user based on vocabulary, re-tweets, and favorited tweets. Predicts score based on 'friends_ids' with a 75 percent accuracy using my network.

Setup:

• If you plan to grab data for more than just a few users, make sure to prevent your machine from sleeping in your energy settings, or find another way to keep the network connection alive.
• Make sure you have tweepy installed
• Before you run this tool, make sure to create/populate a file named 'secret' in the root directory with your authentication keys from twitter (each on a separate line):
  • CONSUMER_KEY
  • CONSUMER_SECRET
  • ACCESS_KEY
  • ACCESS_SECRET

Execution:

1. python download_data.py (leave this running until you have downloaded enough data)
2. python generate_user_scores_dict (check out the histogram of user scores)
3. python predict_scores.py (check out how the predicted scores compare to the actual)

Sources:

positive words:

• http://positivewordsresearch.com/list-of-positive-words/
• http://boostblogtraffic.com/power-words/
• Bing Liu, Minqing Hu and Junsheng Cheng. "Opinion Observer: Analyzing and Comparing Opinions on the Web."

#####negative words:

• https://github.com/jeffreybreen/twitter-sentiment-analysis-tutorial-201107/blob/master/data/opinion-lexicon-English/negative-words.txt
• http://www.slate.com/blogs/lexicon_valley/2013/09/11/top_swear_words_most_popular_curse_words_on_facebook.html