Twitter Trends can be manipulated, says MIT! | socialexcerpts.com

 Rings a bell? No?

Let us tell you a story of step function.

Trending topic on Twitter or Trends are identifed as words with a # in front of them, created organically by the candid Twitverse (Twitter Universe). It is pretty redundant here to mention how important it is to brands and social media agencies who try to be on top of every aspect of the market for their product and services effectively. Trending topics are very dynamic and can help expand ones social base and build a consistent group of followers. So imagine, if you could predict what will be the next popular trend before it happens? It will be sort of like having a time machine.

Professor Devavrat Shah from MIT might have that magic formula. He and a fellow student from MIT have developed an algorithm which can predict those intriguing hashtags up to an hour on average and in some cases, 4-5 hours ahead of the trend phenomena. Cool or what?

There are other types of programs that, in theory, try and perform the same type of function. Twitter has its own algorithm to predict the same but its accuracy is questionable. So what will this algorithm mean to Twitter world? It will mean looking through the Twitter traffic and trying to match up with what seems like a certain model. It can be programmed to look for a certain “step”, as one topic becomes prominent against the general background chatter.

According to Shah, these models are ‘simplistic’ and it is still unknown if trends have a step function that can lead to a correct prediction or any predictions at all, which means a step function is a mathematical function with only a finite amount of pieces.

So what makes Shah’s algorithm so special? It is the fact that this algorithm is trained by Shah and Nikolov using a training set that contained 400 topics, 200 of which trended and 200 didn’t. When left alone to work, the algorithm managed to pull out all the successfully trending topics from the unsuccessful ones and with 95% accuracy and 4% false-positive rate (topics that were predicted to trend but they didn’t).

Shah’s algorithm compares new topic to training set and compares the traffic over the time between them. If this topic develops a similar pattern to one of the successful trends in the training set, it is assigned a weighted number on the possibility of its future success. To sum up, the more a possible trend compares to previous successful trend history, the greater the chance it will be popular. Is it magic or math?

It is very clear that this algorithm is going to be another huge marketing tool. How social media agencies and brands use it, is something we will have to wait for and watch! Do you think this algorithm is going to change the world of social media as we know it? Let us know.