African Bus Routes Redrawn Using Cell-Phone Data
The largest-ever release of mobile-phone data yields a model for fixing bus routes.
Traffic is a major drain on productivity.
Researchers at IBM, using movement data collected from millions of cell-phone users in Ivory Coast in West Africa, have developed a new model for optimizing an urban transportation system.
The IBM model prescribed changes in bus routes around the around Abidjan, the nation’s largest city. These changes—based on people’s movements as discerned from cell-phone records—could, in theory, slash travel times 10 percent.
While the results were preliminary, they point to the new ways that urban planners can use cell-phone data to design infrastructure, says Francesco Calabrese, a researcher at IBM’s research lab in Dublin, and a coauthor of a paper on the work. “This represents a new front with a potentially large impact on improving urban transportation systems,” he says. “People with cell phones can serve as sensors and be the building blocks of development efforts.”
The IBM work was done as part of a research challenge dubbed Data for Development, in which the telecom giant Orange released 2.5 billion call records from five million cell-phone users in Ivory Coast. The records were gathered between December 2011 and April 2012. The data release is the largest of its kind ever done. The records were cleaned to prevent anyone identifying the users, but they still include useful information about these users’ movements. The IBM paper is one of scores being aired later this week at a conference at MIT.
The IBM work centered on Abidjan, where 539 large buses are supplemented by 5,000 mini-buses and 11,000 shared taxis. The IBM researchers studied call records from about 500,000 phones with data relevant to the commuting question.
Mobility data is created when someone uses a phone for a call or text message. That action is registered on a cell-phone tower and serves as a report on the user’s general location somewhere within the tower’s radius. The person’s movement is then ascertained as the call is transferred to a new tower or when a new call is made that connects to a different tower.
While the data is rough—and of course not everyone on a bus has a phone or is using it—routes can be gleaned by noting the sequence of connections. And IBM and other groups have found that these mobile phone “traces” are accurate enough to serve as a guide to larger population movements for applications such as epidemiology and transportation (see “Big Data from Cheap Phones.”)
Cell-phone data promises to be a boon for many industries. Other research groups are using similar data sets to develop credit histories based on a person’s movements and phone-based transactions, to detect emerging ethnic conflicts, and to predict where people will go after a natural disaster to better serve them when one strikes.
To do such tasks in the developing world, there may be little or no other data to work with. Owners of smartphones that have GPS can allow apps like Google Maps to use their location data for traffic sensing information shared with others. But location information on the simple phones that are far more prevalent in the developing world is known only to the mobile carriers. And that data is available only by special arrangement with the carriers.
In the case of transportation, improving roads and public transit systems often depends on labor-intensive work such as the traveler surveys done commonly in the rich world. “The cost of traditional surveys is very high for developing world applications, but cell-phone use is high, so cell traces are a terrific data opportunity. This is a valuable investigative effort,” says Kara Kockelman, a transportation researcher at the University of Texas, Austin.
Indeed, while in a number of past studies mobile phone data was used to infer travel routes and demand, IBM says this was the first time such data was used in an effort to actually optimize a city transit network.
IBM calls its model AllAboard. For Abidjan, the model selected among 65 possible improvements to conclude that adding two routes and extending an existing one would do the most to optimize the system, with a 10 percent time savings for commuters.
Of course, unclogging one transportation route can have unanticipated problems, like attracting more people to use that route, perpetuating the problem. “If travel times noticeably fall on many roadways, many travelers may shift back to peak times” and popular roadways, Kockelman says.
Still, if the data were available in real-time—rather than months after it was created—the results could be even more powerful. “This would provide snapshots of people moving around in a city, allowing the optimal shifting of routes, and reducing travel and wait times,” Calabrese says.
Keep up with the latest in Big data at EmTech Digital.
The Countdown has begun.
March 25-26, 2019
San Francisco, CA