top of page
20180409 DataCity new logo2-09.png


Predicting traffic congestion

How can road data be used in combination with other open data to improve traffic predictions in real time?


Congestion is an issue that affects all drivers and road operators. Every day in Île-de-France, there are several hundred kilometres of traffic jams, wasting commuters' travel time and raising pollution levels. Gaining greater awareness of how traffic jams are formed, particularly in unusual circumstances such as outside rush hours, could help road operators reduce the impact of congestion via: alternative routes, staggered departures, and more reliable travel timings.


154 hours

per year spent in traffic jams

by motorists in Île-de-France



The Île-de-France Highways Department (DiRIF) and the Setec research and design department decided to explore potential solutions in collaboration with the startup Quantcube, which specialises in artificial intelligence for predictive solutions on a micro and macro scale. The experimentation attempted to see if it was possible to predict unusual congestion scenarios (outside rush hour) and, in more general terms, whether algorithms could be used to predict traffic effectively. The partners therefore worked together to develop a predictive model in real time based on several data sources: Floating Car Data (FCD) providing vehicle speeds, but also weather data, and data relating to public holidays, etc. in order to anticipate the formation of traffic jams more accurately and as far upstream as possible.



Quantcube Technology is a startup specialising in artificial intelligence and real-time analysis of massive unstructured data to predict macroeconomic and financial trends such as economic growth, inflation and countries’ political stability. In order to achieve this, the team has developed algorithms which notably can monitor shipping globally and process satellite images automatically.





Several traffic congestion estimation algorithms were developed during the experimentation based on data collected in the public space. The results were compared to reality then linked to improve accuracy. This challenge is the first step towards a more extensive development in the coming years, where working methods for traffic management will be based on more intensive use of predictive algorithms and larger volumes of data (meteorological readings, incident management, etc.).



bottom of page