2016 INFORMS Annual Meeting Program

91 / 561

INFORMS Nashville – 2016
91
2 - A General Approximation Method For Bicriteria
Minimization Problems
Stefan Ruzika, Department of Mathematics/Natural Sciences
University of Koblenz-Landau Universitätsstraße 1 56070
Koblenz (Germany),
We present a general technique for approximating bicriteria minimization
problems with positive-valued, polynomially computable objective functions.
Given 0 < ≤ 1 and a polynomial-time -approximation algorithm for the
corresponding weighted sum problem, we show how to obtain a bicriteria ( · (1 +
2 ), · (1 + 2/ ))-approximation algorithm for the budget-constrained problem
whose running time is polynomial in the encoding length of the input and linear
in 1/ . Moreover, we show that our method can be extended to compute an ( ·
(1 + 2 ), · (1 + 2/ ))-approximate Pareto curve under the same assumptions.
3 - Bicriteria Analysis Of The Fixed-charge Network Flow Problem -
Separating Fixed Costs And Flow Costs
Michael Stiglmayr, University of Wuppertal, Wuppertal, Germany,
The fixed-charge network flow problem is an inherently biobjective optimization
problem: Minimize fixed (design) costs and minimize flow costs. In its classical
form the sum of these two objectives is minimized which corresponds to the
weighted-sum scalarization of the associated biobjective problem.
However, design costs and flow costs are not directly comparable, since design
costs occur once, while flow costs are due periodically. In this talk we present
heuristic and exact solution approaches based on the two-phase method and
ranking algorithms.
4 - Multi-objective Optimization Of Coupled Systems
George Fadel, Mechanical Engineering Department Fluor Daniel
Engineering Building Clemson University, Clemson SC 29634
USA,
An engineering problem consists of two multi-objective problems that must be
coordinated. The top level focuses on the optimal placement of components
under the hood of a car, with design variables which specify the location of the
various non-convex components in a non-convex volume, and non-overlap
constraints. Then, the optimization of shape and size of a battery pack that is one
of the components placed under the hood is conducted. We show how the two
problems can be assigned to separate teams, and their optimizations can be
coordinated, enabling the chief designer to allow the sub-problem or the upper
level design team to be driving the solution.
SC65
Mockingbird 1- Omni
Machine Learning, Big Data and Economics
Sponsored: Information Systems
Sponsored Session
Chair: Beibei Li, Carnegie Mellon University, Heinz College, Pittsburgh,
PA, 15213, United States,
1 - Modeling User Engagement In Mobile Content Consumption With
Tapstream Data
Yingjie Zhang, Carnegie Mellon University,
Low engagement rate and high attrition rate have been major challenges for the
success of mobile apps. To date, little is known towards how companies can
improve user engagement and business revenues through designing effective in-
app pricing strategies. We propose a structural model by accounting for
time-varying nature of engagement and consumer forward-looking behavior. We
analyze mobile tapstream data from a popular mobile reading app. Our results
enable us to tailor optimal pricing strategy to each consumer based on their
engagement status. Interestingly, we found such engagement-specific pricing
strategy leads to lower average price for consumers and higher overall business
revenues.
2 - Examine Large Scale App Usage Structure By Graphical Model
Jinyang zheng, University of Washington, Seattle,
Identifying an app which generates usage to other app(s) is not only a crucial task
for industrial practitioner, but also challenging for researchers given the larger
scale of App network. With a state of art graphical model method, we overcome
the limitation of traditional econometric causal inference models and examine the
causal relationship in emerging App market among Chinese users. Our model
generates a causal diagram displaying usages of what app would that of each
specific app leads to. Spillover effects of certain app and sequential causal effects
can be easily identified, suggesting significant role of graphical model in business
analytics and big data related research.
3 - Airbnb And Hotel Latent Quality
Uttara Ananthakrishnan, Carnegie Mellon University,
Sharing economy has empowered consumers to communicate their needs with
one another and thus has helped them to assume the role of both suppliers and
producers seamlessly. In this paper, using a natural experiment set up and a novel
dataset, we analyze how Airbnb has impacted the traditional way of conducting
the hotel business. We study if the hotels have responded to the increasing
number of Airbnbs by increasing their quality and whether this response varies
across different types of hotels. We analyze the hotel industry’s response across
different dimensions in quality by not only considering star ratings, but also user
sentiments and latent quality expressed in textual content of reviews.
4 - How Much Is An Image Worth? An Empirical Analysis Of
Property’s Image Aesthetic Quality On Demand At Airbnb
Shunyuan Zhang, Carnegie Mellon University, Pittsburgh, PA,
United States,
Dokyun Lee, Param Vir Singh, Kannan Srinivasan
Sharing economy platforms such as Airbnb are challenged with product quality
uncertainty. To solve the issues, Airbnb has implemented strategies such as
professionally taking high quality photos for hosts and calling them verified. This
paper studies the impact of having verified photos. To assess the aesthetic quality
of images, we use machine learning techniques. Employing Difference-in-
Difference method we find rooms with verified photos are on average 9% more
frequently booked. We separate the effect of photo verification from photo quality
and find an extra $2,455 in yearly earnings brought by high photo quality. We
find asymmetric spillover effects across rooms in the same neighborhood.
SC66
Mockingbird 2- Omni
Additive Manufacturing
Sponsored: Quality, Statistics and Reliability
Sponsored Session
Chair: Prahalad Krishna Rao, Assistant Professor, Binghamton
University, P.O. Box 6000, Binghamton, NY, 13902-6000, United States,
1 - Accelerated Process Optimization For Laser-based Additive
Manufacturing By Leveraging Similar Prior Studies
Amir M Aboutaleb, Mississippi State University, Industrial &
Systems Engineering Department, Mississippi State University, MS,
39762, United States, Linkan Bian
Manufacturing parts with target properties and quality in Laser-Based Additive
Manufactured (LBAM) parts is crucial towards enhancing the “trustworthiness”
of this emerging technology. We propose a novel process optimization method by
directly utilizing experimental data from previous studies as the initial
experimental data to guide the sequential optimization experiments of the
current study. We conduct a real-world case study that optimizes the relative
density of parts manufactured using a Selective Laser Melting system. A
combination of optimal process parameters is achieved within 5 experiments.
2 - Online Detection For Cyber Attacked Additive Manufactured Parts
By Real-time Sensing And Analysis
Chenang Liu, Virginia Tech, Blacksburg, VA, 24061, United States,
, Tomilayo Komolafe, Zhenyu Kong,
Jaime Camelio
Cyber security of additive manufacturing (AM) is important for some critical
applications such as defense industry. This work focuses on the online detection
of attacked AM parts by real-time sensing using network analysis based data
fusion techniques. Using the effective features extracted from multiple sensor
data, the discrepancy between normal and attacked AM parts can be detected
effectively. The case study show that the proposed method can successfully detect
the attacked parts, but does not cause false alarm for the sample normal part.
3 - Laplacian Eigen Compressive Sensing For Dimensional Integrity
Classification In Additive Manufacturing
Prahalad Rao, Binghamton University,
This work relates the effect of parameters, namely, infill and extrusion
temperature in fused filament fabrication (FFF) additive manufacturing (AM)
process on pre-selected geometric dimensioning and tolerancing (GD&T) features.
Next, a method is proposed to classify the part quality in terms of the geometric
integrity using minimal number of laser-scanned point cloud data. The proposed
method combines spectral graph theory with compressive sensing, as a means of
supervised classification of part geometric integrity.
SC66