2016 INFORMS Annual Meeting Program

247 / 561

INFORMS Nashville – 2016
247
2 - On The Optimal Timing Of Meld Score Updates In The Liver
Transplantation System
Sepehr Nemati, Ivey School of Business, 1255 Western Road,
Ivey Business School, London, ON, N6G 0N1, Canada,
Zeynep Icten, Lisa N Maillart,
Andrew J Schaefer, Mark Roberts
Patients on the waiting list for liver transplants have opportunities to game the
system by concealing changes in their health status. We formulate a model that
determines, as a function of the last reported health status, current health status,
days until the next required update and the quality of the current liver offer,
whether the patient should do nothing, report her current health status, or accept
the current liver offer (if any) to maximize expected lifetime. We analyze the
degree to which a patient can benefit from the flexibility inherent to the current
reporting requirements.
3 - Operational And Financial Decisions Within Proportional
Investment Cooperatives
Xiaoyan Qian, PhD Student, University of Auckland, 12 Grafton
Road, Auckland, 1010, New Zealand, ,
Tava Olsen
In a proportional investment co-op, operational and financial decisions are
inseparable because members’ capital investment is required to be in proportion
to their economic transactions with the co-op. In agriculture where yield and
market are uncertain, we propose a Markov decision process wherein the
decisions of processing quantity interact with the financial decisions of retained
earnings and short term loans. The results include: (1) characterization of the
value function and the optimal policy; (2) explicit expressions for the
deterministic-yield dynamic program; and (3) identification of financial risks.
Keywords: cooperative; finance and operations; coordination.
4 - Easy Affine Markov Decision Processes: Applications
And Algorithms
Jie Ning, Case Western Reserve University,
Matthew J Sobel
Affine Markov decision processes (MDPs) with continuous state and action
vectors and decomposable constraints on actions have unique features that free
them from the curse of dimensionality. Exploiting the properties of affine MDPs
(a companion paper presented in another session), we present algorithms that
efficiently compute an optimal policy and the value function. We show that affine
MDPs are applicable in a variety of decision-making contexts such as fishery
management and advertising, and that the optimal policies generate qualitative
insights.
TA41
207C-MCC
Quantitative Methods in Finance VI
Sponsored: Financial Services
Sponsored Session
Chair: Xuefeng Gao, The Chinese University of Hong Kong,
William MWM Engineering Building, Shatin, NA, Hong Kong,
1 - A Primal-dual Iterative Method For Stochastic Dynamic
Programming And Its Applications
Nan Chen, Chinese University of Hong Kong,
Due to the curse of dimensionality, people often rely on computationally
tractable, but suboptimal, heustric policies to solve stochastic dynamic programs
(SDP). Our work develops a recursive approach from the technique of
information relaxation to obtain a sequence of confidence intervals for SDP
optimal value. The width of the confidence interval can be used to measure the
quality of currently used heuristics. We also show the resulting intervals converge
in a finite number of iterations to the true value. Thereby our approach presents a
systematic way to improve the quality of control polices. Two applications in
optimal trading execution and network revenue management are discussed.
2 - Operational Risk Management: Coordinating Capital Investment
And Firm Growth
Lingjiong Zhu, Florida State University,
We consider a jump-diffusion model to analyze the impact on a firm’s value of
small shocks caused by market risk events and large shocks caused by operational
risk events. We consider the investments in the infrastructure of a firm that aims
at mitigating the impact of operational risk events through changes in the
stochastic nature of the large shocks. We study the investment strategies in two
settings: the maximization the firm’s value over a fixed investment horizon and
the minimization of the ruin probability over an infinite horizon. This is based on
the joint work with Yuqian Xu and Mike Pinedo.
3 - Limit Theorems For Hawkes Processes With A Large
Initial Intensity
Xuefeng Gao, The Chinese University of Hong Kong,
Hawkes process is a class of simple point processes that is self-exciting and has
clustering effect. The intensity of this point process depends on its entire past
history. It has wide applications in finance, neuroscience, social networks,
criminology, seismology, and many other fields. In this paper, we study the linear
Hawkes process with an exponential kernel in the asymptotic regime where the
initial intensity of the Hawkes process is large. We derive limit theorems for this
asymptotic regime as well as the regime when both the initial intensity and the
time are large. The limit theorems could be useful for approximating the transient
behavior of Hawkes processes.
4 - Testing The Capital Asset Pricing Model Under Economic
Regime Shifts
Yonggan Zhao, Professor, Dalhousie University, 6100 University
Avenue, Suite 2010, Halifax, NS, B3H 4R2, Canada,
We present a dynamic version of the Capital Asset Pricing Model (CAPM) with
economic regime shifts. Assuming the equilibrium security returns are
characterized by economic indicators, we test the hypotheses that risk premiums
on financial securities are asymmetric across economic regimes with positive risk
premium in the expansion regimes and negative risk premium in the contraction
regimes. Using a sector rotation investment strategy, the superiority of the
dynamic CAPM to the traditional CAPM in predicting stock returns is shown.
TA42
207D-MCC
Choice Models and Assortment Optimization
Sponsored: Revenue Management & Pricing
Sponsored Session
Chair: Sumit Kunnumkal, Indian School of Business, Hyderabad, India,
1 - Assortment, Pricing And Market Expansion
Ruxian Wang, Johns Hopkins Carey Business School, Baltimore,
MD, 21202, United States,
We incorporate market expansion into consumer choice models and investigate
the revenue management problems. We characterize the structure of the optimal
policies for the problems under the consumer choice models with various market
expansion effects, and develop efficient algorithms.
2 - Assortment Planning Decision In Two-sided Market
Ying Cao, University of Texas at Dallas,
Dorothee Honhon, Sridhar Seshadri
We consider a firm which makes product assortment decisions when facing a two-
sided market, which means it receives revenues from two distinct groups. We
obtain structural properties of the optimal assortment and theoretical bounds on
the performance of heuristic policies, showing the value of considering both sides
of the market.
3 - A Near-optimal Exploration-exploitation Approach For
Assortment Selection
Vashist Avadhanula, Columbia University, New York, NY, 10027,
United States, Shipra Agrawal,
Vineet Goyal, Assaf Zeevi
We consider a dynamic assortment optimization problem where customers choose
according to an unknown MNL choice model. In each period, we offer an
assortment of at most K products out of N and observe the customer’s choice to
learn the model parameters. We present an exploration-exploitation policy that
achieves a near-optimal worst case regret of $\tilde {O}(\sqrt{NT})$. Our policy is
based on the principle of optimism under uncertainty and does not require any
separability assumption on the parameters. We also present a nearly matching
lower bound of $\Omega(\sqrt{NT/K})$ for this problem.
4 - New Bounds For Assortment Optimization Under The Nested
Logit Model
Sumit Kunnumkal, Indian School of Business,
We consider the assortment optimization problem under the nested logit choice
model. We establish new bounds on the quality of revenue ordered assortments.
TA42