Rolling Horizon Network Revenue Management Using Decomposition Special case of hotel

In this paper, we develop an approach in network revenue management problem, focused on hotel industry. The objective is to determine the booking price for the next 350 dates (one year) in advance, as well as to update them periodically. The main difficulty with determining the optimal solution of multiple-night stay is caused by the huge number of products. Each product (customer request for booking) depends on many factors such as duration of stay, arrival and departure dates, day of week and season. The key idea in this approach is to decompose the multiple-night stay problem into single-night ones. To do so, we need to estimate the effective arrival rate for each individual date by considering the effects of other dates. In fact, customer loss” of a specified date may be due to the features of other dates. Therefore first we calculate the probability of customer loss, which can be categorized into two groups: (a) because of room shortage of at least one date of customer request; and (b) because of high price. By calculating the probability of these events and their effect on the potential customer appearance rates, we estimate the effective arrival rate of each date individually. We develop an algorithm to estimate the effective arrival rate of each individual date based on historical data base. At last, we develop a single leg rolling horizon model to obtain the suitable booking price in a timely manner, within the framework of dynamic programming. By sequentially solving the developed models, an appropriate booking price of different dates are obtained.

Keywords:

revenue management , dynamic pricing , network management , hotel industry