Optimization of public transport after analyzing GSM data

A group of researchers from IBM, using information on the movements of 500 thousand mobile users, have developed a model for improving public transport routes.

The model was successfully tested in the city of Abidjan (population 3.8 million people, the capital of Côte d'Ivoire). The illustration shows the existing bus routes (crimson) and improvements offered by the algorithm (blue). He proposed 65 possible improvements, including three new routes, with an overall time saving of 10% in passenger-minutes for all passengers, taking into account travel time and waiting time.

The new algorithm is useful not only for optimization, but also for laying new routes in newly built neighborhoods. Here he clearly shows in what ways, at what time and in what volume the human mass moves. You can make the best routes and transport schedule. What is important is the fact of the emergence of a new tool that can be used by the city authorities to optimize the infrastructure of the metropolis.


85 SOTRA public transport routes in Abidjan
')
Each owner of a cell phone acts as an individual sensor in the network. Information from the sensors is collected by the cellular operator and analyzed. In this case, information on calls in the Abidjan mobile network from December 2011 to April 2012 was provided by Orange. The base includes 2.5 billion records and today is the largest database of this kind available for scientific research. Naturally, the base is cleared of any personal information: all statistics are anonymous.


Density of users, by place of residence (left) and place of work (right)

In Abidjan, the transport link consists of 539 buses in the SOTRA network, 5,000 minibuses and 11,000 public taxis. The authors of the scientific work used information about calls and SMS from 500 thousand mobile phones. During each call, the operator saves information about the base station that serves the subscriber, which allows determining his coordinates with sufficient accuracy. Movement of the phone is registered if another cell begins to service it later.


Comparison of time before and after optimization of SOTRA routes, in hundreds of thousands of passenger-minutes for the entire passenger traffic

In general, such “surveillance” in real time can be very useful. For example, even crime prediction algorithms based on information from the cellular network are being created . If this IBM algorithm is applied not on historical data, but on real-time information, then theoretically it is even possible to promptly edit the public transport timetable, instantly reacting to an increase in human traffic along certain routes.

The data processing algorithm called AllAboard was made up of specialists from the Dublin laboratory of IBM Research, participating in the Data for Development program. The final report “AllAboard: a system for exploring urban mobility and public transport using cellphone data” was presented at the NetMob 2013 conference, which is devoted to processing databases from cellular networks.


Average waiting time at stops (above) and the density of passenger traffic along different routes (below)

You can get acquainted with the scientific work in the collection of conference materials (pdf) , pp. 397–411.