2016 INFORMS Annual Meeting Program

176 / 561

INFORMS Nashville – 2016
176
Inference For The Progressively Type-I Censored Step-stress
ALT Under Interval Monitoring
David Han, University of Texas at San Antonio, Management
Science & Statistics, College of Business, San Antonio, TX, 78249-
0632, United States,
Tianyu Bai
A step-stress ALT under progressive Type-I censoring was considered when a
continuous monitoring of failures is infeasible but inspections at certain time
points is possible. In addition to the accelerated failure time model, a general scale
family of distributions was considered for flexible modeling. The MLE of scale
parameters and their conditional densities could be derived explicitly. Using the
exact distributions of the estimators, confidence intervals for the parameters were
obtained and numerically assessed.
A Probabilistic Unit Commitment Model
Kenneth Bruninx, KU Leuven, Celestijnenlaan 300, Post Box
2421, Leuven, B3001, Belgium,
, Erik Delarue
Stochastic unit commitment models allow calculating an optimal trade-off
between the cost of scheduling and activating reserves, load shedding and
curtailment, but may become computationally intractable for real-life power
systems. Therefore, we develop a probabilistic unit commitment (PUC)
formulation, which allows internalizing the reserve sizing and allocation in a
deterministic unit commitment problem, considering the full cost of reserve
allocation and activation. This PUC formulation yields UC schedules that are
nearly as cost-effective as the theoretical optimal solution of the stochastic model
in calculation times similar to that of a deterministic equivalent.
Understanding Vehicle Movement Patterns With Artificial
Neural Networks
Burak Cankaya, Lamar University, 13960 Hillcroft St, Apt 724,
Houston, TX, 77085, United States,
This research evaluates the question if we can understand the vehicle movement
patterns depending o their (Geographic Identification Systems)GIS data. The
research propose to classify the vehicle movements as the work they are doing
with machine learning algorithms. The results label the vehicle movement states
and make it possible to evaluate the performance of the vehicles which is an
essential need for work vehicles, personal devices and vehicles, and vessels. The
poster also explains the data preparation and the algorithms such as decision tree,
random forest, and neural networks to classify the geospatial data.
Algorithms For Identifying Optimal Inspection Paths In
Pipe Networks
Thomas Ying-Jeh Chen, University of Michigan, 1780 Broadway
Street, Apt S128, Ann Arbor, MI, 48105, United States,
, Seth Guikema
The inspection of aging water distribution pipes is an important process for
utilities. Due to limitations on inspection capabilities (~2% length of system is
typically inspected annually), an optimization process is needed to suggest
inspection paths. This paper examines the use of 3 algorithms (Genetic Algorithm,
Simulated Annealing, Greedy Search) in finding paths that maximizes high risk
pipes being inspected while reducing the number of pipe feature changes
(material, diameter etc.). The algorithms were applied to a grid network and a
virtual water distribution network. Both cases demonstrated genetic algorithms
were the most effective in identifying strong candidates for inspection.
Properties Of Location Based Social Networks And Travelers
Destination Choice
Ying Chen, Research Assistant Professor, Northwestern University,
600 Foster St, Evanston, IL, 60208, United States,
, Hani S. Mahmassani, Fei Zhao
The aim of this study is to investigate the relationship between friendship and
distance, the possible influence of friendship in travelers’ destination choices, and
the importance of this factor in choosing a destination. By analyzing social
network properties of two Location based Social Networks (LBSNs), the
characteristics of LBSNs are identified. Results show that in general, the distance
has the strongest influence on travelers’ destination choices, followed by personal
preference and social influence from their friends on-line. For users whose friends
are in geographical proximity to each other, a possible synchronization
characteristic amongst individuals is investigated.
Stadiums And Contraband: A Study On Metal Detectors In
The Field
Nelson Christie, Rutgers University, Princeton, NJ, United States,
Sports stadiums are increasingly using walk-through metal detectors for patron
screening. We utilized experimental design to understand detection rates of real
contraband items. These items were identified through interviews of various
subject matter experts. Experiments were carried out on machines borrowed from
stadium venues. We also created a testing scheme for the metal detectors to
ensure functionality prior to events.
Assessing Uncertainty: A Model-output Oriented Approach
Achim Czerny, Dr, Hong Kong Polytechnic University,
Hong Kong, Hong Kong,
Erik T. Verhoef, Anming Zhang
The present paper develops the concept of continuous uncertainty types, which
are defined by the extent to which uncertainty affects the firm’s optimized price
markups and quantities (i.e., “model outputs”). We show that this model-output
orientation can cover scenarios where additive, multiplicative and many more
stochastic structures all occur with positive probabilities. This approach allows a
compact assessment of the impacts of uncertainty. We further show that the
optimal inventory level, and the composition of inventory in terms of the number
and size of production units, depend strongly on the type of uncertainty and its
distribution as defined according to our theory.
Explaining Energy Bonds’ Option-adjusted Spread (OAS) Using
Multiple Exponential Regression Models
Yan Deng, PhD Student, Cornell University, Cornell University,
2406 Hasbrouck Apartment, Ithaca, NY, 14850, United States,
In order to explain the OAS of corporate energy bond, we developed exponential
regression models for prediction. We found that as the oil prices drop, the OAS of
energy bonds widen significantly. The sensitivity of OAS to oil price varied among
energy subsectors in line with leverage. In addition, adding treasuries yield
predictor could significantly increase the predicting accuracy. In particular, these
two variables can explain 87.4%, 75.6%, 86.5%, 64% and 84.4% of credit spread
changes for independent, integrated, midstream, oil field, and refining energy
bonds respectively. Our predictive model provided a tool to monitor risk and
signal rich or cheap bonds as potential buy/sell candidates.
Deep Learning For Sleep Assessment
Skyler C Devine, University of Tennessee - Knoxville,
Knoxville, TN, 37916, United States,
Based on the physiological and neurological features, sleep is divided into two
main types: Rapid Eye Movement, and non-rapid eye movement. NREM sleep
consists of three stages, stages 1-3. Brain activity during sleep stochastically
alternates between stages. In order to judge sleep, clinicians record the electrical
activity of the brain through an electroencephalogram, and visually inspect the
results to classify them into the three stages. This process is referred to as “sleep
scoring”. We apply a deep learning algorithm to automatically score sleep and
provide monitoring of sleep quality.
Inventory Placement Supply Chain With Two
Competing Retailers
Yi Ding, Southeast University, Sipailou 2, Jiangsu Province,
Nanjing, 210096, China,
This study examines service time competition in the context of inventory and
environmental constraints. We first discuss the case of a downstream duopoly
market without regulator, and then we extend the model by including a regulator
that is dedicated to carbon emission abatement. We analyse how service time can
be affected internally through inventory placement and externally through
market competition as well as government regulation of carbon emissions. The
results suggest that although expedited service requires higher safety stock,
increasing unit inventory holding cost does not seem to slow down service, nor
does imposing higher carbon tax.
Decision Support Model To Planning A Mobility Scheme For
Critical System Services In Urban Networks With Natural
Interruptions
Andrea Margarita Ditta, Universidad del Norte, km 5 Antigua vía
Puerto Colombia, Barranquilla, Colombia,
Ruben Yie, Gina Galindo
This work aims to design a Decision Support Model (DSM) to planning a mobility
scheme in emergency scenarios in urban networks. The DSM seeks to evaluate
the transportation between points of incidents and points of care. The research is
focused in the area of humanitarian logistics considering natural interruptions like
streams, storms, downpours, among others. We undertake emergency response
systems with critical services. Fire brigade, police force requirements or urgent
medical attention, are examples of critical hope to increase the
efficiency in dealing with emergencies, by decreasing attention times and risk of
accidents.
Supermarket Optimization: Simulation Modeling And Analysis Of
A Grocery Store Layout
Jessica Peggy Dorismond, University at Buffalo, 3028 Elmwood
Avenue, Buffalo, NY, 14217, United States,
This is a study on how to optimize the layout of a supermarket in order to
increase its gross profit via the maximization of impulse sales. In most
supermarkets many items often get unnoticed because on average customers only
walk one-third of the store. Recent advances in marketing research reveal that
encouraging customers to walk longer paths can often increase spending because
they are exposed to more products. Retailers can then increase their sales by
using the store layout—i.e., the design of the aisles and the product location—to
extend the customers’ shopping paths and thus indirectly motivate them to
purchase items that are not originally on their shopping list.
POSTER SESSION