INFORMS Nashville – 2016
417
2 - Prescriptive Analytics In Airline Operations: Cost Index
Optimization Through Improved Arrival Time Prediction
Anna Achenbach, PhD Candidate & Research Assistant, WHU -
Otto Beisheim School of Management, Burgplatz 2, Vallendar,
56179, Germany,
anna.achenbach@whu.eduFacing enduring cost pressure, the European airline industry has turned to
machine learning to enhance their operations. In this paper, we develop a
prescriptive model which improves the prediction of an aircraft’s flight time
taking as inputs different speed levels, flight and weather information. Through a
more accurate flight time prediction an optimal cost index can be determined by
considering increased fuel burn against cost of time. The study is based on
random forest regression combined with convex optimization of the cost index
using actual flight data of a European airline.
3 - Adaptation And Relational Contracting In The US Airline Industry
Ricard Gil, Johns Hopkins Carey Business School, 100
International Drive, Baltimore, MD, 21202, United States,
ricard.gil@jhu.edu, Myongjin Kim, Giorgio Zanarone
In the airline industry, ex-post adaptation of flight schedules is necessary in the
presence of bad weather. When major carriers contract with independent
regionals, conflicts over adaptation decisions typically arise. This paper analyzes
the importance of relational adaptation in the airline industry. Our model shows
that the long-term value of the major-regional relationship must be at least as
large as the regional’s total cost of adaptation across joint routes. Thus, the major
is more likely to preserve routes outsourced to regional airlines that have higher
adaptation costs. Using the 2008 financial crisis as exogenous shock, we find
consistent evidence with our theoretical predictions.
WB63
Cumberland 5- Omni
Location Models in Humanitarian Logistics
Sponsored: Location Analysis
Sponsored Session
Chair: Kayse Lee Maass, University of Michigan, 1205 Beal Avenue,
Ann Arbor, MI, 48109-2117, United States,
Leekayse@umich.edu1 - Emergency Shelter Location Analysis And Logistics For
Hurricane Preparedness
Nicholas E. Lownes, University of Connecticut,
nlownes@engr.uconn.eduDeterministic and stochastic location models are presented as decision support
tools for emergency shelter planning in the state of Connecticut. Several scenarios
are evaluated, including local and statewide evacuation strategies. Jurisdictional
boundary and backup power generation constraints are also examined in the
context of evacuation strategy.
2 - Perishable Commodity Allocation In Humanitarian Supply Chains
With Multiple Demand Types
Melih Celik, Middle East Technical University,
cmelih@metu.edu.tr, Ozlem Ergun, Julie L Swann
In this study, we consider the allocation of a perishable item in a humanitarian
supply chain, which consists of a capacitated warehouse and distribution locations
(DLs). Each DL serves multiple demand classes with varying benefits for satisfying
the uncertain demand, and applies a usage policy (prioritization of the higher
classes or first-come, first-served). We develop algorithms that maximize the
expected net benefit of the system. When demand arrivals are Poisson, we
establish structural results on relative performance of usage policies and
centralization of the DLs. These results are extended using computational
experiments based on the H1N1 vaccination campaign in the state of Georgia.
3 - Time-variant Adaptive Robust Optimization For
Hurricane Preparedness
Xinfang Wang, Georgia Southern University, Statesboro, GA,
30460, United States,
xfwang@georgiasouthern.edu, Jomon A Paul
We propose an adaptive robust model for determining the stockpile location and
capacities and impacts of these decisions on social costs (logistic, deprivation and
fatality), given time-variant hurricane characteristics. Uncertainty set size is
adjusted according to storm advisory released every six hours, making the model
adaptive to time-variant uncertainty realization.
WB64
Cumberland 6- Omni
Representations and Approximations in
Multiobjective Optimization
Sponsored: Multiple Criteria Decision Making
Sponsored Session
Chair: Serpil Sayin, Koc University, Istanbul, Turkey,
ssayin@ku.edu.tr1 - On Measures Of Quality For Representations And Approximations
Serpil Sayin, Koc University,
ssayin@ku.edu.trWe revisit the definition of discrete representations of the nondominated set in
multiobjective optimization. We focus on the coverage error as a measure of
quality, review some related measures and investigate relationships among them.
We note that the computational features of a measure become critical while
building algorithms that deliver representations that conform to given quality
expectations. We report on our computational experience with a lower bound on
the coverage error as embedded in a recent algorithm that delivers
representations. Finally we analyze the output of our algorithm in relation to an
approximation of the nondominated set.
2 - On The Existence Of Ideal Solutions In Multi-objective Binary
Programming And Its Applications: Preprocessing, Valid
Inequalities, And Probing
Hadi Charkhgard, University of South Florida, Tampa, FL,
United States,
hadi.charkhgard@gmail.com, Natashia Boland,
Martin Savelsbergh
Studying objective functions of multi-objective binary problems (MBPs) can be
helpful in detecting MBPs that can be solved in polynomial time. This talk shows
that in addition to this important fact, studying this topic can also result in
developing some preprocessing, valid inequalities, and probing techniques that
can be incorporated in any solver of MBPs. More precisely, this talk focuses on
studying objective functions to show that under which conditions a MBP has an
ideal solution. Based on that, it shows that (1) how the number of objectives may
be reduced; (2) how some classes of valid inequalities may be generated; and (3)
how the size of a MBP may be reduced by eliminating some certain variables.
3 - A Novel Multi-criteria Optimization Method For Volumetric
Modulated Arc Therapy (vmat) Treatment Planning
Gokhan Kirlik, University of Maryland School of Medicine,
Baltimore, MD, 21201, United States,
gokhankirlik@umm.edu,Warren D. D’Souza, Hao Howard Zhang
Volumetric modulated arc therapy (VMAT) is one of the radiation therapy
techniques for cancer patients. In VMAT, the main goal is to deliver right amount
of radiation dose to the tumor while sparing the normal tissue. Therefore, VMAT
requires considering several conflicting objectives, which is called multi-criteria
optimization (MCO). In MCO, efficient solutions are used instead of the optimal
solution. In this study, we use achievement scalarization to obtain efficient
solutions for VMAT. We tested our approach on 10 locally advanced head-and-
neck cancer cases. We demonstrated that the proposed MCO method was able to
obtain VMAT plans with significant improvement in dosimetric plan quality.
WB65
Mockingbird 1- Omni
Digital Innovation & Analytics
Sponsored: Information Systems
Sponsored Session
Chair: Eun Ju Jung, George Mason University, George Mason
University, Fairfax, VA, 22030, United States,
ejung6@gmu.edu1 - Promotion Strategy, Operating Mechanism, And Economic Value
Of Digital Platform For Consulting Services: A Two-sided
Market Perspective
Tae Hun Kim, Michigan State University, East Lansing, MI, United
States,
thkim@broad.msu.edu,Kyung Ho Song, Daegon Cho
A platform plays as a two-sided market by matching clients with consultants in a
digitized way for consultancy. To show its value, we focus on promotion strategy,
operating mechanism, and economic impact. The success depends on its attraction
for potential customers initially and IT-mediated operation for existing clients
over time. Its economic value is explained for both client and consultant sides. We
first explore possible ways to promote the digitized services. Second, specific IT
features are identified to enable user engagement and interactions. Finally,
economic value is analytically proved by a pricing model for the two-sided market
and a synergy between online and offline services.
WB65