Deriving all Passenger Flows in a Railway Network from Ticket Sales Data

In optimizing the national railway timetable for the Belgian infrastructure company Infrabel, we use the quality criterium of minimizing expected total passenger travel time. Optimizing the timing of all trains fairly over all passengers requires passenger flow information. Focus is usually on determining the train service offer side, being the train timing, without considering the demand side data, being the passenger flow information. In this paper, we focus on passenger flow derivation. We start from train ticket sales data and use Dijkstra’s modified shortest path algorithm on the Belgian passenger train graph to determine a route per passenger origin destination station pair. We sum route passenger flows over all these routes. We also show how, in optimization of a timetable, timing and flows should be treated as interdependent. This means that we can use a process of iteration over these two phases, which we then call reflowing and retiming. The infrastructure company did not have this passenger flow information available for the current schedule, but they need it to optimize many decisions. We determined these passenger flows for the current timetable. We also give some recommendations regarding the improvement of passenger flow related data recording procedures, which would significantly improve passenger service.