2016 INFORMS Annual Meeting Program

260 / 561

INFORMS Nashville – 2016
260
TA76
Legends D- Omni
Decision Analysis
Contributed Session
Chair: Jordi Weiss, PhD Student, Unil, Lausanne, 1022, Switzerland,
1 - Multi-modal Optimization Of WINTIME As A Game Performance
Metric And Rankings Basis With An Application To
College Football
Christopher Keller, Assistant Professor, East Carolina University,
Department of Marketing & Supply Chain, College of Business,
Greenville, NC, 27858-4353, United States,
WINTIME is the elapsed clock time for the winning team’s score to exceed the
losing team’s final score. WINTIMES can be used to generate a rating system and
an estimated WINTIME. The resulting optimization problem is to minimize the
errors between the observed and the estimated WINTIMEs. For college football,
the model has 129 variables and is multi-modal. Excel Solver solutions are
discussed. Accuracy is comparable to other systems with predictive accuracy
above 70% and retrodictive accuracy above 85%. The system could also be
applied to other sports like hockey or soccer.
2 - Should I Stay Or Should I Go? The Cognition Of Exploration
And Exploitation
S.S. Levine, University of Texas, Dallas, TX, 75080, United States,
, Charlotte Reypens
In many life situations, people choose sequentially between repeating a past
action in expectation of a familiar outcome (exploitation), or choosing a novel
action whose outcome is largely uncertain (exploration). For instance, in each
quarter, a manager can budget advertising for an existing product, earning a
predictable boost in sales. Or she can spend to develop a completely new product,
whose prospects are more ambiguous. Using experiments in a lab and a labor
market, we examine what affects these decisions. We investigate traits of the
decision-makers, such as risk aversion, but also their history. We find that the past
matters, greatly: What you experience counts as much as who you are.
3 - A POMDP Model For Personalized Depression Monitoring
Jue Gong, Graduate Student, University of Washington, Industrial
& Systems Engineering, Box 352650, Seattle, WA, 98195,
United States, , Shan Liu
Mitigating depression has become a national health priority as it affects 1 out of
10 American adults. We formulate a partially observable Markov decision process
(POMDP) in order to find an optimal monitoring schedule for anindividual
patient. The state of the POMDP combines the health state of the patient and the
direction of health change. We estimated the transition and emission matrices by
extending the Baum-Welch algorithm to include a mixture of multiple transition
matrices. We solved the model using the Bellman Equation via dynamic
programming algorithm.
4 - Remanufacturing Decisions In A Close-loop Supply Chain With
Extended Warranty Options
Kunpeng Li, Utah State University, 767 Eagle View Dr.,
Providence, UT, 84332, United States, ,
Yang Li
The study addresses the problem of choosing the appropriate reverse channel
structure for collecting of used products. We consider a two-echelon supply chain
with a single manufacturer and a single retailer. By comparing five different
reverse channel formats, we intend to understand how close-loop supply chain
structure influences the use-product collection. We also study the impact of
extended warranty on the consumption of remanufactured products.
5 - Using Online Games To Develop Manager Intuition About
Demand Randomness
Jordi Weiss, PhD Student, Unil, Lausanne, 1022, Switzerland,
, Michael Bean, Suzanne de Treville
Quantitative-finance methods applied to the supply chain dramatically improve
incorporation of demand risk into supply-chain decisions. Use of these methods is
hindered by managers lack of intuition about demand randomness. We use online
games to allow managers to apply such methods in the face of randomness
arriving from different forecast-evolution processes (instantaneous volatility or
jump diffusion). We present the results from using these games at the policy level
for three cantons in Switzerland, demonstrating how increased intuition about
randomness helps decision makers to consider profitable options that are
counterintuitive and non linear.
TA77
Legends E- Omni
Opt, Integer Programing I
Contributed Session
Chair: Ed Klotz, IBM, PO Box 4670, Incline Village, NV, 89450,
United States,
1 - Solving Large Scale Grid-based Location Problems
Noor E Alam, MD, Assistant Professor, Northeastern University,
334 SN, 360 Huntington Avenue, Boston, MA, 02115,
United States, , John Doucette
This talk will present mathematical models for grid-based location problems
(GBLPs) with two case studies. Apart from presenting computational difficulty of
the GBLPs, it will also discuss two problem-specific integer linear program (ILP)
based decomposition algorithms to solve large-scale instances.
2 - A Mixed-integer Programming Approach To Optimize Typing
Method And Design Of Touchscreen Keyboards On Smartphones
Mohammad Ali Alamdar Yazdi, PhD Student, Auburn University,
354 W Glenn Ave, Auburn, AL, 36830, United States,
, Ashkan Negahban, Fadel Mounir Megahed
Millions of people use smartphones for different typing purposes. There are
significant differences in the design of keyboards for smartphones. Optimization
of typing method and improvements in the design of touchscreen keyboards on
smartphones is expected to have a significant impact on the total typing time. A
MIP model is developed to optimize typing zones for fingers in two-thumb typing
with the objective to minimize the total typing time. Through extensive
experimentation, different keyboard dimensions are also compared to find the
optimal keyboard design. The results shows that the best keyboard design has
square keys with minimum possible horizontal and vertical spaces between the
keys.
3 - Solution Value Contingent Cuts For Solving Hard Generalized
Assignment Problems
Robert M Nauss, Professor, University of Missouri - St Louis,
3816 Boca Pointe Dr., Sarasota, FL, 34238, United States,
Jeremy William North
Define hard generalized assignment problems (GAP) to be those that take more
than one hour CPU time to prove optimality. Tremendous strides have been made
in the capability of “off the shelf” software, such as GUROBI, to solve general
integer linear programs (ILP). However, some classes of ILPs remain difficult to
solve to optimality in a reasonable amount of time. Certain instances of the GAP
exhibit this behavior. While good feasible solutions are found relatively quickly,
the issue remains in proving optimality. We introduce some novel cuts (and
methods for deriving said cuts) that are applied with an “off the shelf” solver
through the use of CALLBACK functions. Computational results are presented.
4 - Improved Analysis Of Infeasible Mixed-integer Linear And
Quadratic Programs
Ed Klotz, IBM, P.O. Box 4670, Incline Village, NV, 89450, United
States, John W Chinneck, Andrew Scherr
Analysis of infeasible MILPs and MIQPs is complicated by the integer restrictions
(IRs). Current techniques return a minimal infeasible subset of the linear
constraints and variable bounds by solving a series of MILPs. They do not find a
minimal subset of the IRs, because of the significant additional computational
cost. We develop efficient ways to find a minimal subset of the IRs and use this to
speed the isolation of a true Irreducible Infeasible Subset for MILPs and MIQPs.
This is helpful when a variable is accidentally specified as integer.
TA78
Legends F- Omni
Opt, Network I
Contributed Session
Chair: Seyed Mohammad Nourbakhsh, Sabre, 3150 Sabre Drive,
Southlake, TX, 76092, United States,
1 - Enhanced Robust Operational Aircraft Routing
Seyed Mohammad Nourbakhsh, Sabre, 3150 Sabre Dr, Southlake,
TX, 76092, United States, ,
Dong Liang, Xiaodong Luo, Xiaoqing Sun, Sergey Shebalov
Enhanced Robust Operational Aircraft Routing (E-ROAR) serves as an airline
planning decision support solution that is used as a post-process to the Fleet
Assignment Model (FAM) to bridge the chasm between Airline Planning and
Airline Operations. E-ROAR considers connections, and maintenance
requirements; and provides fleet assignments and through connections feasible
for Crew Planning, Maintenance Planning, and Airline Operations. The output
from E-ROAR is given to Airline Operations, where tail assignments are made.
We propose an advanced solution approach by combining heuristic and
optimization methodologies. Computational results demonstrate significant
improvement.
TA76