The new approach to “friend suggestions” looks at the usual haunts of individuals to determine which users may have connections with one another. — File Photo

LONDON: A method of predicting which individuals may become friends on social networking sites based on the places they visit out in the real world has been developed by researchers at Cambridge University in Britain.

The new approach to “friend suggestions” looks at the usual haunts of individuals to determine which users may have connections with one another.

This, combined with the “friend-of-a-friend” method, currently favoured by social networking sites such as Facebook and LinkedIn, can increase the efficacy of the prediction system, say researchers.

“We wanted to investigate the properties of places that encourage connections between visitors and how this could be incorporated into a system that predicts friends,” Salvatore Scellato, one of the researchers, told Reuters.

The team analysed the creation of social connections on Gowalla, a location-based social networking site that allows users to share information about the places they visit.

“We monitored the behaviour of people going to places and the connections they made. We found that lots of people who go to the same places end up adding each other as friends, accounting for around 30 per cent of new social links,” Scellato said.

Meeting places are given incremental value based on how likely they are to foster connections and interaction between people, therefore offices, gyms and schools carry more weight than football stadiums or airports.

The weighting was determined according to the number people who visit the place, and the regularity of those visits -- a theory Scellato describes as “place entropy”.

“We considered the entropy of a place to find venues that are more likely to foster social links, such as offices and gyms, rather than train stations or museums. We discovered that two users visiting a place with low entropy, that is, a place with a handful of people who are regulars, are highly likely to develop a social connection,” said Scellato.

Friend prediction systems pose a large problem to social networking sites due to the vast numbers of people they deal with. Facebook has more than 750 million active users.

“Our results show it's possible to improve the performance of link prediction systems on location-based services that can be employed to keep the users of social networks interested and engaged with that particular website,” Scellato said.