Exact algorithms for continuous pricing with advanced discrete choice demand models
Name
291_2024_Article_799.pdf
Size
1.97 MB
Format
Adobe PDF
Checksum (MD5)
30a0625f4011bdd1d220dea0c3e218c9
Author(s)
Haering, Tom
Legault, Robin
Torres, Fabian
Ljubić, Ivana
Bierlaire, Michel
Date Issued
November 26, 2024
Journal
OR Spectrum
Publisher
Springer Berlin Heidelberg
Citation
Haering, T., Legault, R., Torres, F. et al. Exact algorithms for continuous pricing with advanced discrete choice demand models. OR Spectrum (2024).
Version
Final published version
Abstract
We present a spatial Branch and Bound and spatial Branch and Benders Decomposition approach together with the Breakpoint Exact Algorithm (BEA) to tackle the uncapacitated choice-based pricing problem (CPP) where demand is captured by a discrete choice model (DCM) based on the random utility principle. We leverage problem characteristics to reformulate the state-of-the-art simulation-based formulation of the CPP as a mixed-integer linear program (MILP) into a non-convex quadratically constrained quadratic program (QCQP), and then into a non-convex QCQP with linear objective (QCQP-L). We solve this reformulation with an efficient spatial Branch and Bound procedure utilizing the McCormick envelope for relaxations, which are then solved using Benders decomposition. We further exploit utility breakpoints to develop the BEA, which scales polynomially in the number of customers and draws, providing a fast option for low numbers of prices. Our methods are evaluated against solving the MILP, QCQP, or QCQP-L with GUROBI on a mixed logit (ML) parking space operator case study. We outspeed the MILP by several orders of magnitude when optimizing one or two prices and reduce computational time drastically for larger numbers of prices. When comparing to algorithms tailored for the CPP with ML demand specifically, our approaches significantly outperform the state of the art. Our methodology suits all choice-based optimization problems with linear-in-price utilities, given any DCM.
MIT Department
Massachusetts Institute of Technology. Operations Research Center
Terms of Use
Creative Commons Attribution
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.1007/s00291-024-00799-3
