Robust Optimization of Seat Allocation for High-Speed Railways Using Re-Identified Passenger Demand

With the aim of optimizing seat allocation for high-speed railways, this paper applies machine learning techniques to re-identify the “true” willingness of ticket-purchasing based on real-world ticketing data. Using the re-identified demand, uncertainty demand is formulated, and a robust optimization approach is proposed to better address the uncertainty in passenger demand by introducing a control parameter. A series of experiments with different parameter values are designed for measuring the performance of the robust model. Results show that the proposed approach outperforms the state-of-the-practice by around 4.8% on ticket revenue as well as increases the load factor rate by around 3.2%.