Data-driven joint optimization of pricing and seat allocation in trains

Abstract

Revenue management (RM), a management science method, employs demand predictions to maximize revenue. Since its introduction in the airline industry in the 1980s, it is now widely used in various industries such as hotel, retail, and railway, among others. Compared to the airline industry, penetration has been slow in the railway industry. In recent years, RM has seen increased adoption in the railway industry. Existing approaches to RM in trains have mostly considered the determination of price and capacity separately. In this paper, we consider the problem of maximizing revenue by optimizing over price and capacity simultaneously. We make three important contributions to this problem. For the first time, we consider the problem of train sizing, i.e. number of coaches in each fare class. Second, instead of logit-based models estimated from price-demand data, we consider data-driven models that alleviate the problem of modelling errors. Finally, we formulate the optimization as an integer linear program as opposed to the nonlinear formulations that result from logit-based models. Thus, we present data-driven revenue maximization to jointly determine pricing, seat allocation, train sizing for a single fare class over multiple legs and train lines. Our simulation results targeting Nozomi on the Tokaido-Sanyo Shinkansen in Japan show that it is possible to formulate a railway operation plan to maximize the operation profit by applying our proposed method. In our numerical experiments, the proposed approach can identify solutions that increase profit per day by 50%, with a computation time of less than 1 second.