EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
2655. Dynamic Pricing of Empty Seats in the Transportation Industry under the Nested Logit Model
Invited abstract in session TB-59: Pricing and Capacity Management, stream Pricing and Revenue Management.
Tuesday, 10:30-12:00Room: S08 (building: 101)
Authors (first author is the speaker)
| 1. | Davina Rauhaus
|
| Mercator School of Management, University of Duisburg-Essen | |
| 2. | Christiane Barz
|
| Institute of Business Administration, University of Zurich | |
| 3. | Jochen Gönsch
|
| Mercator School of Management, University of Duisburg-Essen |
Abstract
We suggest a dynamic pricing model for selling "empty seats" - seat reservations for extra space in addition to regular reservations. Such extra space tickets share the main product’s resources and are a significant revenue-generating opportunity when coaches, trains, or airplanes frequently depart with many empty seats. We formulate the problem of a transportation company that sells tickets in the same compartment (1) without a seat reservation, (2) with a seat reservation, and (3) with a seat reservation and extra space as a Markov decision process. To address the resulting curse of dimensionality, we follow two approaches. First, we use approximate linear programming (ALP) with an affine value function approximation. We solve the resulting linear problem via a column generation algorithm. The result can be shown to provide both a tighter bound than previously known as well as a simple heuristic for the control problem. To allow for more general basis functions and possibly good value function approximations and corresponding heuristics, we also discuss simulation-based approximate dynamic programming (sb-ADP). In a numerical study, we show that our sb-ADP heuristic typically obtains expected revenues close to our theoretical upper bound and better than a simple ALP-based policy. The quality of the sb-ADP, however, highly depends on the allotted computation time. For increasingly complex customer preferences, the time required for strong performance can be prohibitive.
Keywords
- Revenue Management and Pricing
- Complexity and Approximation
- Programming, Dynamic
Status: accepted
Back to the list of papers