The customized bus system is an innovative demand-responsive public transit service with the potential to significantly alleviate congestion and environmental footprint. To fully exploit the flexibility of this approach, it is pivotal to forecast the demand for the service, in order to optimize the use of vehicles and resources. In this paper, with the aim for supporting the use of customized bus systems, we formalize the predictive task and assess the performance of a range of machine learning techniques. We introduce a two-step predictive task aiming at (i) identifying the presence of demand and, if there is actual demand, (ii) estimating the number of passengers to be served. The experimental analysis, based on realistic data from the Beijing area, shed some light into the performance of different classes of approaches.
Period6 Sep 2023
Event title11th Symposium of the European Association for Research in Transportation
Event typeConference
Conference number11
LocationZurich, SwitzerlandShow on map
Degree of RecognitionInternational