Joint MIT/INFORMS Symposium
Sunday, April 25, 2004
1:00 - 1:45 When OR is AND
Thomas L. Magnanti, Dean of MIT School of Engineering
1:45- 2:20 Overview of the Operations Research Center
James B. Orlin and John N. Tsitsiklis, Codirectors
B R E A K
2:50- 4:10 Supply Chain Management
Modeling and Optimization of Supply Chains: Opportunities,
Experiences and Challenges - Stephen C. Graves, MIT and
Sean Willems, Boston University
We will present a modeling framework for optimizing the placement of
inventory across a supply chain. We will describe the implementation
of the model into software and examples of its application to practice.
We discuss our experiences and learnings from our efforts to disseminate
this research to practice, and conclude with opportunities and challenges
based on these experiences.
Inventory Inaccuracy: Its Causes, Consequences, and Cures
– Yua Kang, MIT and Stanley Gershwin, MIT
Inventory management is central in Operations Research. Nearly all
the authors of the technical literature, however, make one reasonable-sounding
false assumption: inventory managers know exactly what they are storing.
In reality, inventory records are only estimates, and often not good
estimates. Recent technology promises to make inventory estimates far
more accurate. This talk describes inventory inaccuracy and ways to
compensate for it.
A Portfolio Approach for Procurement Contracts - David
We focus on the procurement of commodity products, such as electricity,
steel, grain, cotton or computer memory, which are typically available
from a large number of suppliers each of which offers different contracts
at a different price and a different level of flexibility. Despite the
non-strategic nature of commodity products, procurement decisions need
to take into account inventory risk associated with inventory shortages
or unsold products, as well as price risk associated with spot markets.
We show how a portfolio approach, that has been applied recently by
a number of manufacturing companies, can be used to find the right trade-off
between expected profit and risk.
2:50-4:10 Data Driven Models
Clustering Models to Improve Forecasts in Retail Merchandising
- Nitin Patel, MIT and Cytel Software Corp., Mahesh Kumar, MIT,
and Rama Ramakrishnan, ProfitLogic, Inc.
Forecasting short-term demand for hundreds of items is a routine activity
in retail merchandising. Methods used in practice employ a parametric
demand model for each item. Our approach clusters items into homogenous
groups and models parameter estimates for each item as following a multivariate
normal distribution. We will describe two clustering algorithms based
on this approach. We will report on experience with applying these algorithms
to several real and simulated data sets.
Mining Customer Quality of Service Measures from Transactional
Data - Les D. Servi*
Some data rich systems measure transactional data (what happens when?)
but cannot directly measure quality of service indicators (such as customer
delay) needed for capacity investment decisions. This talk summarizes
an approach to infer QoS measures as well as discusses its actual implementation
in the field.
* The author currently works at MIT Lincoln
Laboratory. This work was conducted while the author was working
at Verizon Laboratories and is based on joint work with Daryl Daley
(Australian National University).
Pattern Classification and Machine Learning via Large-Scale
Optimization Methods - Robert M. Freund, MIT and Ryan Rifkin, Honda Research Institute USA, Inc.
Pattern classification and machine learning methods are used in a wide
variety of settings: to distinguish benign from malignant tumors in
cancer diagnosis, or to automatically detect objects in the path of
a high-speed vehicle for accident avoidance. In this brief talk we will
review the key modeling tools used to develop pattern classification
and machine learning algorithms using large-scale optimization methods.
B R E A K
4:40 - 6:00 OR in the Public Sector
Airline Security: A Lost Cause? - Arnold I. Barnett,
Airline security has improved since 9/11, but the overall record is
not stellar. Indeed, several anti-terrorist measures that could readily
pass cost/benefit tests have actually been abolished since the catastrophe.
We discuss recent developments in passenger and luggage-hold screening,
and the potential value of passenger profiling systems.
25 Years of Urban Operations Research - Richard
C. Larson, MIT
Cities experience the full gamut of OR problems. This talk will review
the author's experience doing OR in New York City, in writing the textbook
(with Amedeo Odoni) called Urban Operations Research, and in supervising
MIT ORC thesis students on urban OR problems. We conclude with an update
-- a new book (with Amedeo Odoni) is in the works, the Internet is helping
to get the word out thanks to MIT's OpenCourseWare (OCW) initiative,
and -- thanks to targeted support from the Lounsbery Foundation -- we
are establishing a global on-line community of practitioners and scholars
in Urban OR.
Network Economics in an Emerging Economy: The Case Study of
India - Vijay Chandru, Indian Institute of Science and
Wireless telecommunications technologies like GSM, CDMA, WLL (wireless
local loop) are having a dramatic impact on the economics of communication,
information transfer and markets in countries like India. Close to a
100,000 new telephone connections are wired every day and the per minute
cost of telephony (local and long distance) are the lowest in the world.
This talk will present an analysis of this data in the context of models
of economic growth.
4:40 - 6:00 Financial Engineering and Revenue Management
The Lifetime Optimizer - Dimitris J. Bertsimas,
MIT and Gina Mourtzinou, American Express
We describe an application of robust multiperiod optimization to plan
the financial life of an individual or a family. The optimizer is the
main engine in American Express financial planning software that American
Express recently launched in collaboration with Morningstar to serve
the needs of its 11,000 advisors and 2.5 million clients.
Manufacturing Revenue Management: A Negotiation Support System
for Make-To-Order Transactions - Jérémie
Gallien, Yann Le Tallec and Tor Schoenmeyr, MIT
Our objective is to identify near-optimal dynamic guidelines for the
sales negotiation of price, quantity and delivery lead-time terms in
make-to-order contracts. We study accordingly the dynamic programming
formulation of a related admission control problem, derive a computationally
efficient policy using a fluid approximation, and report its performance
in numerical experiments. Finally, we discuss the potential implementation
of this work into a distributed decision support system.
Revenue Management: Where Have We Been, Where Are We Going?
- Barry Smith, Sabre Holdings
Revenue Management is the practice of managing price and availability
of products or services with the goal to maximize revenue or profit.
Since deregulation of the US domestic airline industry, revenue management
has evolved from a fringe to mainstream business function and has contributed
billions of dollars in incremental profitability. The concepts and tools
are now used across a wide range of industries. We’ll review the
origins of revenue management, current practices and future directions.
4:40 - 6:00 Some Realities of Decision Making
Model-based Rationality - Jeremy F. Shapiro, MIT
and Slim Technologies
At the dawn of the information age, von Neumann and Morgenstern posited
economic principles of rational decision-making that have evolved over
the years in parallel with developments in information technology. Organizational
behaviorists have found these principles to be unrealistic as the result
of studying how decisions in organizations are actually made in the
face of ignorance, ambiguity and conflict. In this talk, we present
a new set of principles, called model-based rationality, describing
how managers can, should and, ultimately, will exploit data-driven,
descriptive and prescriptive models to improve their decision-making.
The Practice of Decision and Risk Analysis - Samuel
Bodily, University of Virginia
A review of contemporary practice and challenges in decision analysis
and risk management will touch a variety of topics: (i) strategic decision
analysis, (ii) structured risk management, (iii) real options and hedging,
(iv) approximations of probability distributions, (v) decision trees
and influence diagrams, (vi) spreadsheet decision modeling and risk
analysis simulation, and (vii) decisions involving multiple attributes
related to time and stakeholder outcomes.
Human Health Risks of Animal Antibiotics - Louis
Anthony (Tony) Cox, Jr., Cox Associates and NetAdvantage, Inc.
Food safety regulators worldwide fear that antibiotics used in food
animals select for antibiotic-resistant bacteria that infect people
and cause excess illness-days and fatalities as human antibiotic treatments
fail. Europe has already banned key animal antibiotics. Yet, quantitative
data and OR modeling show that these feared risks are tiny compared
to the risks from not using animal antibiotics to prevent bacterial
diseases. We consider how to regulate animal antibiotics more rationally
to protect human health.