Project Mentor: Stan Nikolov
The goal of this project is to “Tweetstrap”/bootstrap a marketer to find the twitter mini-celebrities (beyond the “usual suspects”) to target to help market his test tweet. The predictions are done based on the influence source of the tweet i.e. popularity of the source tweeter, the length of the cascade chains that the tweet generated and also analysis of the content of the tweets (topics from the words or links it contains). The aim is to develop a regression model (support vector based) that will predict the popularity of a tweet (by giving retweet counts) , factoring in various attributes of tweets from a training set for each celebrity user. We aim to use a corpus of tweets extending across mini-celebrity twitter users from various domains, in order to get a good distribution of topical events, that they have been interested in(tweeting about) and consistent distribution of tweets from “twitter celebrities” who are usually retweeted.
Command line way to get sample predictions : python predict_cli.py “Test tweet string”
App link : http://tweetstrap.kuldeepk.com