MIT boffin's algorithm foresees Twitter's trendy future
Predicts with 95% accuracy when a small trend will burst big
An MIT professor has developed an algorithm to predict what's going to be big on Twitter before it's big on Twitter – for those that care about such things. Twitter will care, he predicts.
Associate Professor Devavrat Shah and his student Stanislav Nikolov have created an algorithm that determines with 95% accuracy whether a topic will trend. Shah can find out up to four or five hours before Twitter lists it on their "Trending Topics". On average, he can predict whether topics such as #youknowyouredumpedwhen, #christmas, #itsfriday will trend 90 mintues before they make it onto Twitter's trending list.
It's partly to do with a pattern of "explosive growth" early on, but even that doesn't fully predict whether something goes "a step up" and hits the big time, or whether it will die off. If the pattern of growth matches that of other Twitter topics that proceeded to go big, they are weighed more heavily in the counting.
The researchers' results, reported in MIT News, are based on a survey of 400 twitter topics, of which 200 trended and 200 didn't.
Shah said that the data sort of worked out what was going to be big by itself:
Trending things ... remain small for some time and then there is a step. This is a very simplistic model. Now, based on the data, you try to train for when the jump happens, and how much of a jump happens.
The problem with this is, I don't know that things that trend have a step function. There are a thousand things that could happen. So instead [we] just let the data decide.
MIT's newsletter predicts that Twitter may be interested in the widget as it could help them sell and target advertising. ®