2016 INFORMS Annual Meeting Program

488 / 561

INFORMS Nashville – 2016
488
WE14
104D-MCC
Facility Location II
Contributed Session
Chair: Jianmai Shi, Associate Professor, National University of Defense
Technology, Changsha, 410073, China,
1 - The Maximal Covering Location Problem With Minimization Of
Cardinality Of Upper Average
Eric C Blair, PhD Student, University of Florida, Gainesville, FL,
32601, United States, , Matthew Norton
We present a reformulation of the Maximal Covering Location Problem (MCLP)
using the new concept of Cardinality of Upper Average (CUA). In the MCLP, a
fixed number of facilities are located to minimize the number of customers that
must travel farther than some maximum desirable distance to the nearest facility.
In the new formulation utilizing CUA, the number of customers with the largest
travel distances whose average travel distance to the nearest facility is equal to the
maximum desirable distance is minimized. The resulting problem is reduced to a
mixed-integer linear program. We demonstrate the advantages of the new
formulation with a numerical example.
2 - The Optimal Planning Of Electric Vehicle Battery Swapping
Network With Fuzzy Customer Satisfaction
Fang Guo, Huazhong University of Science and Technology, 1037
Luoyu Road, Hongshan District, WuHan,, WuHan, 430074, China,
Key to the mass adoption of electric vehicles is the establishment of sufficient
battery service Infrastructure network based on customer behavior and
psychology. We present an EV battery swapping network planning with fuzzy
customer satisfaction, which aims to determine the location and service capacity
strategy of stations simultaneously with the consideration of customer
satisfaction. A heuristic algorithm is proposed to solve the problem. Furthermore,
we conduct the parameter analysis when EVs are used in the practice of the City
Cluster in the Middle Reaches of the Yangtze River in China.
3 - Logistics Service Network Design For Humanitarian Response In
East Africa
Marie-Eve Rancourt, Université du Québec à Montréal, 2920
Chemin de la tour, CIRRELT, Montreal, QC, H3T 1J4, Canada,
Émilie Dufour, Gilbert Laporte,
Julie Paquette
The United Nations Humanitarian Response Depot (UNHRD) is an important
humanitarian logistics service provider that manages a network of depots. This
research project aims to analyze the potential benefits of adding a regional
distribution center in Kampala, Uganda, to its existing network. To this end, we
used fieldwork, simulation, optimization and statistical analyses to assess the costs
of prepositioning relief items in Kampala and to propose a robust stocking
solution. The UNHRD has already started to implement the solution proposed in
this study, which should result in a mean cost reduction of around 21%.
4 - Joint Deployment Of Electric Car Charging Stations And Servers
On A Road Network
Jianmai Shi, Associate Professor, National University of Defense
Technology, Changsha, 410073, China, , Yue
Wang, Zhong Liu, Yajie Liu
A joint charging station location and charging piles (servers) configuration
problem for electric cars is studied, where the service capacity of the station
depends on the number of servers. The problem is an extension of the Flow-
Capturing location problem, and an integer programming model is proposed to
maximize the overall car flow served, subject to a finite budget. A heuristic
algorithm is developed to solve the problem, and the performance of the
algorithm is tested by networks with different scales. We finally present a case
study based on a practical road network in China.
WE15
104E-MCC
Text Mining II
Sponsored: Artificial Intelligence
Sponsored Session
Chair: Weifeng Li, University of Arizona,
1 - Exploring Scada Devices And Their Vulnerabilities On The Internet
Of Things
Sagar Samtani, The University of Arizona,
Much of modern society is reliant on critical infrastructure. Much of this
infrastructure is controlled and managed by Supervisory Control and Data
Acquisition (SCADA) systems. Recent years has seen an increase in internet
connectivity of SCADA systems. While this has resulted in an increased level of
convenience, it has also opened up the possibilities for devastating cyber-attacks.
This study demonstrates a data and text mining approach to identify SCADA
systems on Internet of Things. We also use state-of-the-art vulnerability
assessment techniques to identify the vulnerabilities of these devices. The results
of this study indicate that many SCADA vendors are vulnerable to various
exploits.
2 - Key Conversation Trends And Patterns About Electronic
Cigarettes On Social Media
Wenli Zhang, University of Arizona, Tucson, AZ, United States,
Sudha Ram”
Electronic cigarettes (e-cig) usage has increased exponentially over the last few
years and are perceived to be safer alternative to cigarettes use. In order to
understand the public health impact of e-cig, a better understanding of
population-wise use patterns, perceptions regarding the use and abuse liability of
e-cig should be developed. However traditional survey is not adequate to get such
information. The research objective of this study is to explore using social media
data to identify key conversations, trends, and patterns about the usage of e-cig
by using natural language processing, word embedding, topic modeling, content
and sentiment analysis, and social network analysis.
3 - Stock Movements Prediction Using Textual And Technical Data
Juxihong Julaiti, The Pennsylvania State University, 445
Waupelani Drive, J01, State College, PA, 16802, United States,
Predicting stock movements is one of the most appeal topics for researchers since
an accurate predictive enables gain wealth, as well as the rapid progress of data
acquisition has made the vast amount of data available. In this paper, we apply
different machine learning techniques to predict the daily stock movement. In
particular, we use the type, title of news articles of the current day that are related
to the interested company from the Wall Street Journal, as well as its stock
volume of the day. The result of predicting Google’s stock changes indicates big
drops or big growths are easier to capture, and the average AUC and accuracy of
the Gaussian Naïve Bays are 0.83 and 0.97.
WE17
105B-MCC
Nonlinear Optimization Algorithms III
Sponsored: Optimization, Nonlinear Programming
Sponsored Session
Chair: Feng Qiang, Argonne National Laboratory, 9700 S Cass Avenue,
Argonne, IL, 60439, United States,
1 - Analysis Of The Proximal Quasi Newton Algorithm In Solving
Convex Composite Problems
Hiva Ghanbari, Lehigh University, Bethlehem, PA, 18015, United
States, , Katya Scheinberg
In this work, we analyze the convergence properties of an inexact proximal
quasi-Newton algorithm to solve composite optimization problems in the case of
strong convexity. We consider solving subproblems to the eps-optimality while an
inexact sufficient decrease condition is checked to be satisfied at each iteration.
Furthermore, we apply the Nesterov’s accelerated scheme to present the
accelerated proximal quasi-Newton algorithm. In addition to the theoretical
properties of the resulting algorithm, the numerical results will be presented.
2 - Parallel Problem Generation For Nonlinear Programming
Problems In Julia
Feng Qiang, Argonne National Laboratory, Lemont, IL, 60439,
United States, Joseph A Huchette, Miles Lubin,
Cosmin Petra
Large scale optimization problems usually have rich structural properties. In this
talk, we present StructJuMP, a parallel modeling environment for structured
optimization problems. Built as a parallel extension to modelling language JuMP,
StructJuMP offers additional syntax to specify blocks of the problems and an MPI-
based parallelization of the model generation. We demonstrate StructJuMP
capabilities for the specification of power grid optimization problems on an HPC
cluster at Argonne National Lab.
3 - Regularized Primal And Dual Methods For Convex Quadratic
Programming
Elizabeth Wong, University of California-San Diego,
We discuss active-set methods for convex quadratic program with general
equality constraints and simple lower bounds on the variables. In the first part of
the talk, two methods are proposed, one primal and one dual. In the second part
of the talk, a primal-dual method is proposed that solves a sequence of quadratic
programs created from the original by simultaneously shifting the simple bound
constraints and adding a penalty term to the objective function. Numerical results
are presented for the combined primal-dual active-set method.
WE14