MIT algorithm can predict how popular a tweet will be
Maximizing retweets of Twitter posts has always been a science to some Twitter users. They try to post at certain times when a lot of their followers are active. Others try to list the right hashtags or mention accounts of important people. Now, there might be a new consideration to take into account.
A Massachusetts Institute of Technology professor has developed a model that predicts the number of retweets in the entire lifetime of a tweet based on the first couple of minutes of retweet activity.
“A lot of people felt that Twitter usage was totally random and unpredictable, that it was all just noise,” said Sloan School of Management assistant professor Tauhid Zaman. “But it turns out that there is a systematic, repeatable behavior that you can model.”
PHOTOS: Tech giants’ new headquarters
Zaman unveiled a website that charts the retweets of about 50 tweets from politicians and celebrities and shows how they match up with predictions. Offering a social media twist on the Ouija board, the website’s called Twoujia: the Retweet Oracle.
Zaman said the research could help advertisers understand the spread of ideas and give users a basis to monetize their tweets.
He worked on the research with Emily B. Fox, Amazon professor of machine learning at the University of Washington, and Eric Bradlow, co-director of the Wharton Customer Analytics Initiative at the University of Pennsylvania. They’ve submitted the study to the Annals of Applied Statistics.
The latest research follows up on Zaman’s previous research on determining someone’s influence on Twitter. He found that during major events, one Twitter user becomes a “Superstar.” Because of the high activity of their followers, these superstars garner substantially more retweets than other users.
ALSO:
Google releases new racing game playable with mobile Chrome browser
Apple’s Tim Cook grilled at All Things Digital conference opening
Facebook is said to end bid to buy map start-up Waze for $1 billion
