Peak activity forecast for urban transport
Our client operates and runs one of Europe’s biggest urban transport networks. As part of an operational improvement and optimisation approach, MP DATA was called upon to design and develop a passenger peak forecasting tool covering the entire network.
Our Data Scientist team designed and implemented several modelling approaches (machine learning and deep learning) to best use peak history data, network topology data and the different exogenous data (weather, calendar, events) to be able to accurately predict future peaks for each node (stations, stops). The approach made it possible to identify the most accurate and effective solution adapted to our client’s technical and operational constraints.
The solution architecture was conceived as a modular micro-service built into the client’s IS infrastructure and offering the different teams several interfaces.
Network nodes are covered by the solution
of overall average prediction errors on D+1
for flow management, the assignment of vehicles and for infrastructure sizing