Close panel

Close panel

Close panel

Close panel

Technology Act. 18 Aug 2017

The algorithm that identifies future trending topic

twitter bbva

Many companies would pay through the nose for a social network fortune teller who could spot trends in social media before anyone else and develop effective strategies to solve potential problems or ensure that they could get the most out of a particular trend whilst it is booming. A fortune teller might be the most exotic approach, but an algorithm is more effective, and an MIT professor has developed just such a tool that can spot the next trending topic.

Testing has already shown 95% efficiency. Professor Devavrat Shah and his student Stanislav Nikolov have developed an algorithm that can not only determine upcoming trends, it can do so sufficiently far in advance to enable decision making. Their algorithm knows what is going to happen an hour and a half before something becomes a trending topic.

The next trend

Achieving this efficiency first requires data collection. Shah and Nikolov's solution analyses huge volumes of data to establish behavioral patterns, whilst being sufficiently flexible to not to make assumptions about what is going to happen. The tool compares tweets in real time as they are published, thus discovering from the past what the future of these conversations will be and, so, the next trend.

The researchers compared 200 topics that did trend and 200 that did not. The tool's accuracy will increase as more topics are followed and more samples incorporated.Although this solution has been tested on Twitter and achieves its best results with social network publications, there is no reason why this should be its only application.

Any information that can change over time (i.e. where the future cannot be predicted with certainty) can be analyzed with the tool developed by these researchers, such as, for example, how long a bus journey might take, ticket sales for a film and maybe even stock market prices.