2016 INFORMS Annual Meeting Program

279 / 561

INFORMS Nashville – 2016
279
3 - Multi-objective Optimization Via Simulation On
Integer-ordered Spaces
Kalyani S Nagaraj, Purdue University, West Lafayette, IN, United
States, , Kyle Cooper, Susan R Hunter
Consider the context of multi-objective optimization via simulation when the
search space is integer-ordered. We propose a framework to efficiently identify
the Pareto set that solves a sequence of stochastically constrained problems (via
the epsilon-constraint method) and is designed for deployment on a parallel
computing platform. We discuss the design principles that make our framework
efficient.
4 - Parallel Empirical Stochastic Branch And Bound
Jie Xu, George Mason University, 4400 University Dr., MS4A6,
Fairfax, VA, 22030, United States, Scott L. Rosen,
Peter Salemi, Sajjad Taghiyeh
In this talk, we show how the Empirical Stochastic Branch and Bound (ESSB)
algorithm, which is an effective globally convergent random search algorithm for
discrete optimization via simulation, can be adapted to a high-performance
computing environment to effectively utilize the power of massive parallelism.
We propose a master-worker structure driven by MITRE’s Goal-Directed Grid-
Enabled Simulation Experimentation Environment. We present numerical
experiments with a benchmark test function and a real-life simulator to test the
scalability of parallel empirical stochastic branch and bound.
TB46
209B-MCC
Choice Modeling and Assortment Optimization
Sponsored: Revenue Management & Pricing
Sponsored Session
Chair: Huseyin Topaloglu, Cornell Tech, 111 8th Avenue, Suite 302,
Ithaca, New York, NY, 10011, United States,
1 - Assortment Optimization Under General Choice
Srikanth Jagabathula, NYU Stern School of Business,
We consider the static assortment optimization problem where the objective is to
determine the profit/revenue maximizing subset of products from a large universe
of products. The product prices are exogenously fixed and the demand follows a
general choice model. The problem in general is NP-hard and the greedy
algorithm has been found to have good practical performance. We study the
performance of a local search heuristic and show that it reaches the optimal
solution for the MNL model and derive approximation guarantees for the random
parameters logit (RPL) and the nested logit (NL) models. Numerically, we show
that the algorithm outperforms existing heuristics in a wide-range of settings.
2 - Robust Assortment Optimization Under The Markov Chain Model
Antoine Desir, Columbia University,
Vineet Goyal, Bo Jiang, Huseyin Topaloglu, Tian Xie, Jiawei Zhang
In this paper, we consider a robust assortment optimization problem under the
Markov Chain model. In that setting, the true parameters of the model are
unknown and belong to some uncertainty set. The goal is to select an assortment
that maximizes the worst-case expected revenue over all parameter values. We
present an efficient algorithm to compute the optimal robust assortment when
the uncertainty set is row-wise. That is naturally the case in many settings. Our
algorithm provides interesting operational insights regarding addressing
uncertainty in the Markov chain model.
3 - On The Structure Of Cardinality-constrained Assortment
Optimization Problems
Louis L Chen, Massachusetts Institute of Technology-ORC,
David Simchi-Levi
Cardinality-Constrained assortment optimization, the problem of offering an
assortment of items of constrained size that will maximize expected revenue, is
generally regarded as a challenging problem. We provide a new perspective to the
structural analysis, one that illuminates the optimality of “greedy solutions.” The
approach reinterprets some known results for standard choice models but also
provides some new ones as well.
4 - Competitive Pricing Under The Markov Chain Choice Model
Huseyin Topaloglu, Cornell Tech, New York, NY, 10011,
United States, James Dong
We consider competitive pricing problems under the Markov chain choice model.
In this choice model, the customer transitions between the products according to
a transition probability matrix. Based on the price of the product she visits, she
decides to purchase the product or not. In our competitive setting, the prices of
the different products visited by the customers are controlled by different firms.
Each firm wants to maximize its own expected revenue. We show that a Nash
equilibrium exists and the equilibrium prices are lower than those charged by a
central planner.
TB47
209C-MCC
Pricing, Promotions and Bundling for
Revenue Management
Sponsored: Revenue Management & Pricing
Sponsored Session
Chair: Pelin Pekgun, University of South Carolina, 1014 Greene Street,
Columbia, SC, 29208, United States,
1 - A Pricing Model To Optimize The Promotions Period In Airlines
Daniel Felipe Otero Leon, Lecturer, Universidad de los Andes,
Bogota, 1111, Colombia,
Cristina Lopez, Mariana Escallon, Raha Akhavan-Tabatabaei
Promotions help increment the demand for a flight. Several decisions have to be
made to offer a promotion such as its price and duration. We propose a method to
estimate the behavior of customer inter-arrival time distribution, his buying
probability distribution, and the dilution effect from data and develop a stochastic
dynamic model to maximize the revenue, evaluating the decision of whether or
not to offer the promotion. Finally we study the structural properties of the model
and draw conclusions.
2 - Dynamic Pricing For Hotel Rooms When Customers Request
Multiple-day Stays
Yun Fong Lim, Singapore Management University,
Selvaprabu Nadarajah, Qing Ding
We study the dynamic pricing problem faced by a hotel that maximizes expected
revenue from a single type of rooms. Demand for the rooms is stochastic and
non-stationary. Our Markov decision process formulation of this problem
determines the optimal booking price of rooms (resources) for each individual
day, while considering the availability of room capacity throughout the multiple-
day stays (products) requested by customers. To offer attractive average daily
prices, the hotel should not only substantially raise the booking prices for some
high-demand days, but also significantly lower the booking prices for the low-
demand days that are immediately adjacent to these high-demand days.
3 - On The Benefit (or Cost) Of Large-scale Bundling
Tarek Abdallah, New York University,
We study the effectiveness of a simple bundling mechanism in extracting the
consumer surplus in the presence of non-negative marginal costs and correlated
valuations. We develop simple robust analytics that identify the main drivers for
the effectiveness of the pure bundling mechanism and allow the sellers to easily
quantify the potential profits of a large-scale bundling mechanism relative to
more complicated selling mechanisms. Our numerical simulations show that
these analytics provide high predictive power for the true performance of the
bundling mechanism and are robust to different parametric assumptions even for
relatively small bundles.
4 - How Perceptions Of User Reviews Impact Price Competition
Pelin Pekgun, University of South Carolina, Columbia, SC, 29208,
United States, , Michael Galbreth,
Bikram Ghosh
We analyze the interaction of user reviews and experience uncertainty, where
negative and positive reviews may be weighted differently in a consumer’s
assessment of the valence of the posted reviews. We find that the competitive
impact of this unequal weighting may not be intuitive in terms of pricing and
profits. In particular, if consumer awareness is higher for the lower quality
product, it can charge higher prices and realize higher profits in equilibrium than
its higher quality competitor when consumers are strongly influenced by negative
reviews.
TB47