Abstract
This study investigates the revenue management in hotel industry. We focus on the stochastic problem of allocating finite hotel rooms to reservation requests of non-homogeneous Poisson processes. In the hotel industry, customers with different check-in days, stay lengths, room preferences, and valuations arrive randomly during a planning horizon. This study proposes a simulated expected revenue gap decision procedure (SERG) to help hoteliers make an accept-or-reject decision promptly each time a customer arrives and maximize the revenue. We consider three problem complexities by varying the assumptions on the number of room classes and the number of price classes. Also, in the assumption of multiple room classes, we consider whether or not a customer is allowed to change room during the stay. Thus, two optimal models and two versions of SERG are proposed to handle these two cases. In our simulation experiments, under the assumptions of one room class and one price class, the performance of SERG is very close to the optimal solution under perfect information and the gap is 0.21%. Under the assumptions of one room class and multiple price classes, the gap is 0.94%. When there are multiple room classes, the gap in the case of changing room and fixed room are 1.35% and 1.81%, respectively. The average gap between SERG and optimal solution increases as the problem becomes more complex. However, generally speaking, SERG has impressive and robust performances in the above problem environments. Also, it takes only a few seconds to compute the most difficult problem in the experiments; thus, it is feasible to use such an approach in hotel reservation system.