Connexxion aimed to operate a more flexible bus schedule in the future. At times too few vehicles are deployed during a busy period, while at other times too many vehicles are deployed with respect to the occupation. Hence they wanted to find out how you can predict the number of passengers in such a way that they won’t deploy too many buses at quiet times, but also won’t deploy too few buses at busier times.
Capacity Prediction Model
Industry, Data science
The first design started with the route between den Bosch and Eindhoven Airport. With an eXtreme Gradient Boosting prediction model we looked at the recent history (3 months). In the case of holidays, for example, we also looked at the 3 months before the start of the holiday. Then we looked at the historical occupation combined with different factors like the weather, events, and flight data. Using this we made a prediction of the expected demand for the vehicles. After this we looked at the capacity of the vehicles to check if there would be any conflicts.
Dashboard design by Connexxion
For the Proof of Concept we made a prediction model which Connexxion displayed in a PowerBI dashboard. Using these predictions, warnings were created with respect to the capacity, such that Connexxion could anticipate conflicts when making their bus schedules.