Traffic forecast in urban transport
Our client operates on one of the largest urban transport networks in Europe. In a mindset of upgrading and optimizing operations, MP DATA was asked to design and develop a forecasting system in order to predict the number of travelers on the entire transporting network.
Our Data Science team designed and implemented several modeling approaches (machine learning and deep learning) in order to make the best use of traffic history, network topology data and various exogenous insights (weather, calendar, events), allowing us to accurately predict future affluence for each station.
Using this approach, we were able to identify the most adapted solution to our clients technical and operational constraints. Its architecture was a modular microservice integrated into the customer's IS infrastructure thus offering several interfaces for the different teams.
Network nodes were affected by the solution
Prediction error on forecasts on day plus one
Of the flow management, vehicle allocation and infrastructure sizing