Une cartographie prédictive des accidents

1

Business context:— Marseille wants to exploit its datasets and assess potential impact of Big Data Approach for its territory

(Agglomeration)— ENGIE R&D Department wants to show its expertise in big data and machine learning solutions based

on real issues.

Business need:— Big Data / Methodology and Process.

— Proof of concept : Provide Decision tools, Analysis tools for “Big” Event Organization and Mobility in urban areas.

Actions:— The import and processing of heterogeneous data from Marseille IS or Opendata :

weather data, victim’s number of accidents, road characteristics, etc.— Modelization and implementation of a machine learning algorithm to predict the probability of accidents

occurrence and the optimal positions of city resources (police, firemen, etc.)

Results :— 76 % of prediction accuracy in term of accidents occurrence according to a given scenario (weather

and time information)

PREDICTIVE ROADS ACCIDENT MAPTo enhance the roads safety

02/2016 CRIGEN

202/2016 CRIGEN

PREDICTIVE ROADS ACCIDENT MAPTo enhance the roads safety

Perspectives:— Take into account additional information such as the traffic importance, big events (football match,

festival, etc.) and holidays when learning the prediction model— Simulate in real time the cars traffic