E-News Blog
The challenges of keeping America safe have stimulated research in the areas of defense resource allocation and threat detection. At the last day of the 2011 INFORMS conference, three doctoral students at various universities around the nation presented cutting-edge research within the security field.
Jason Yates at Texas A&M is working on a method to approximate the famous network interdiction problem with a less computationally intensive model. The network interdiction problem models sequential adversarial tactical choices made by a hypothetical enemy and the defender of a network. The enemy tries to maximize flow through the network with capacitated links while an interdictor tries to minimize this maximum flow by stopping these flows using resources that are finite. Using real-world networks, Yates provides a method for approximating this intractable problem with an easier equivalent that sacrifices very little in terms of solution quality.
Sinan Tas at the University of Wisconsin-Madison has extended the general framework of previous network protection models to include cascading failures and restoration times of besieged network components. His research investigates the effectiveness of combinations of defense strategies in an electrical power network with the possibility of cascading failures. As part of the effort, he has developed a heuristic method for assessing the vulnerability of such networks.
On the more proactive side of the equation, Christie Nelson of Rutgers University is working on complex statistical learning and testing techniques to filter measured data from radioactive isotopes in order to minimize the chances of failing to identify incoming nuclear material at sea ports. In an unstable world with growing security threats, the role of operations research in understanding and strengthening public safety cannot be overstated.
The research presented in Monday's Quality, Statistics and Reliability (QSR) poster session showed that quality and reliability has moved far beyond manufacturing. The research of 10 presenters from around the world ranged from theory of experimental design to applications in disaster management and healthcare.
PhD student Okan Pala from the University of North Carolina Charlotte presented his research on multi-attribute decision analysis (MADA) for decision makers involved in the recovery of critical infrastructure, such as power grids, after a disaster. Through simulation of various scenarios, he created a simple lookup table to inform decision makers based on the attributes of a particular event. Shahrzad Faghih Roohi from the National University of Singapore used the Bayesian network to guide sensor deployment that balances competing objects of cost and reliability. Virginia Tech PhD student Lee Wells has developed a framework to quickly visualize variation in automobile assembly and identify needed process improvements.
Patrick Wanko of North Carolina Agricultural and Technical State University presented a variation on genetic algorithms dubbed "Schooling Genetic Algorithms." Wanko's research follows the successful trend of nature-inspired metaheuristics by considering how schooling fish move between feeding zones. Marilia Perez from the University of Puerto Rico at Mayagüez displayed her comparison of regression models frequently used in simulation metamodels with kriging, a technique recently borrowed from spatial statistics. Purdue University PhD student Dadi Xing shared his application of variable selection with LASSO to supersaturated experimental design.
The QSR poster session exhibited the breadth of quality and reliability research as it penetrates the complex systems of many fields. Researchers are combining relatively recent developments in statistics, such as LASSO and Bayesian networks, with developments in computing to help businesses and communities increase quality and better serve their markets.
As we've seen, the classical definition of analytics breaks down the category into three: descriptive (or what happened/is happening), predictive (what will/might happen), and prescriptive (what to do to get the desired outcome). Whereas predictive and prescriptive analytics fall under the category of advanced analytics and require expert knowledge to develop reliably, descriptive analytics are probably more prevalent than we know. Anyone who has ever analyzed a data set using Excel or looked at plots to try to determine trends and characteristics has done descriptive analytics of some sort.
However, there are more sophisticated and interesting tools to make sense out of data, and the business intelligence community is promoting their use extensively. On Wednesday morning, Ugo Sglavo from the Advanced Analytics and Optimization Services R&D group at SAS gave a brief overview of the different tools that are available for users interested in transforming data into information. Beginning his talk with the definition of the term "apophenia," which represents the technical term describing seeing meaningful patterns in the data, he gave a plethora of examples ranging from marketing, finance, voice recognition, and text analysis applications to illustrate the different techniques one might use to describe the patterns in the data.
He talked about visualization as a tool to characterize products for up to five dimensions; network graphics to develop better, more meaningful ways of representing graphs and networks; segmentation clustering as one can apply in marketing to group customers; associations such as those used by Amazon to perform market basket analysis; and text analysis as the "dark matter" of IT. Sglavo provided a good overview of the different tools available to perform better descriptive analytics and amass the hidden value of the data we are beginning to be obsessed with gathering.
Applications of operations research to the energy industry are more numerous than ever. From accounting for smart grid technology infrastructure costs in electricity generation expansion planning to reallocating current resources to decrease the operating costs of the power network, applications are plentiful. We've seen some great achievements through the Edelman Prize awarded to the Midwest Independent Transmission System Operator (MISO) team for the use of operations research to optimize the efficiency of the wholesale power market in the Midwest. Continuing on this trend, a joint venture at the Arizona State University (ASU) between the Departments of Electrical Engineering and Industrial Engineering is working on determining better ways to partition the reserve zones to ensure a high reliability for the grid, while awaiting for new technologies to enter the space of power generation.
In technical terms, reserves represent proxy reliability constraints where, although not necessarily ensuring reliability, they provide some flexibility and enable the system to get backed up with minimum disruptions. However, despite their crucial role in ensuring the good functioning of the power system, methods for determining what levels reserves should have and where geographically they should be located are still relatively simplistic. The ASU group is looking at the unit-commitment problem of scheduling generator operating times and output levels to determine where the reserves should be placed to ensure reliability while avoiding additional burden on the network.
Using a mixed-integer formulation, classic network flow models, and connectivity constraints within each zone, the group aims to reduce the computational complexity of stochastic unit commitment by decomposing the problem into two stages, with a deterministic phase followed by a stochastic analysis over multiple possible scenarios. Preliminary results using the IEEE Reliability Test System suggest some decrease in the cost of unreliability while clearly showing that enforcing connectivity within a zone is quite expensive.
As SAS's Keith Collins discussed in Monday's plenary, if businesses want to remain competitive in a modern era, they must learn to make sense of the growing amounts of data available to them. Universities are recognizing that turning data into decisions requires a particular skill set compassing the fields of computing, business, statistics, and management science. Michael Gorman, President of the Analytics section at INFORMS, led a panel discussion with faculty from four schools offering master's degrees in analytics.
Michael Rappa shared his experience with the Institute for Advanced Analytics, which he started at North Carolina State University (NCSU) in 2007. The University of Tennessee, represented by Kenneth Gilbert, introduced a similar program in 2010. Helmut Schneider spoke about Louisiana State University's master's degree, currently in pilot, and Diego Klabjan discussed the program at Northwestern University, which debuts next year. Characteristic of each program is foundational instruction on understanding businesses problems, formulating and solving data-driven statistical models, and communicating results. The speakers shared a desire to break free from traditional academic methods and instead focus on pedagogical methods that could prepare students for real-life problems. Although the programs teach students how to solve statistical problem with software (largely SAS products), the panelists emphasized the importance of teaching students to work in teams and communicate with nontechnical persons.
INFORMS' recent focus on analytics points toward the need for programs such as these, a point only confirmed by NCSU's decision to double its class size for 2012. Analytics and analytics degrees do not replace operations research. These programs will equip graduates with distinctive skills that will work well alongside operations research in helping organizations make better decisions in a complex world.
We want your videos taken at the INFORMS Annual Meeting. There is still time - the deadline is December 9. All suitable videos will be posted on the Annual Meeting website and on the INFORMS YouTube channel. The best video will win an Apple iPad2! Show us INFORMS in your eyes and help us build our video library. For more information, all contest rules, and where to upload your video, visit here.
Coverage of key sessions was provided by the following student reporters for the Annual Meeting eNews Daily. Our thanks goes out to them!
Ana-Iulia Alexandrescu is a graduate student in the Industrial and Systems Engineering Department at Lehigh University. Born and raised in Bucharest, Romania, she came to Lehigh as a Bostiber Scholar in 2006, where she enrolled in the Integrated Business and Engineering Honors Program. She graduated with honors in 2010 and continued on as a Presidential Scholar, pursuing her MS degree in industrial and systems engineering. Outside the classroom, she has been involved in numerous student organizations and initiatives, continuously seeking to improve the quality of student life and to give back to the community. As a graduate student, she held the position of Vice President of the INFORMS Student Chapter at Lehigh for one year, and she is now acting as an advisor to the current executive board of the chapter.
Brendan Bettinger is a doctoral student in the Healthcare Systems Engineering program at Northeastern University, where he is researching applications of game theory in the healthcare industry. He can be reached at Twitter: @ColliderBrendan.
Belleh Fontem (originally from Cameroon) is a doctoral student in operations management at the University of Alabama. He holds a bachelor's in computer science and communications engineering from Duisburg-Essen University (graduating as one of the top students in his class) and a master's in industrial engineering from Cambridge University. Before beginning his doctoral studies in the United States, he did consulting work for both private and public sector institutions in Africa and Europe. He loves debating and salsa dancing, and he is fluent in English, French, and German.
Tim Hopper is a second-year PhD student in operations research at North Carolina State University. His research interests include stochastic optimization, healthcare applications, and automation. Charlotte 2011 was his first INFORMS Annual Meeting. He can be found on Twitter at @tdhopper.
Selected out of six finalists to win the 2011 Daniel H. Wagner Prize, the Decision and Engineering team at Intel, represented by members Evan Rash and Karl Kempf, explained the nature of the problem they attacked and how their solution approach expanded the body of operations research knowledge. Since Intel's founding in 1968, the chip design and manufacturing company has expanded to become a multinational corporation with over $129 billion in market capitalization, serving multiple geographic markets around the world.
Over the years, this explosion in size had created a complex profitability problem along the dimensions of time, markets, and product features.
Because individual markets had unique product feature requirements, Intel's engineers had to find the sweet spot between designing a universal chip for all possible markets (which would be too cumbersome a chip) or customizing the properties of each chip to fit each market (thereby running into an engineering capacity bottleneck).
Reaching an optimal trade-off implied fine-tuning the engineering design process so as to know whether or not to produce a specific chip, which features to include on that chip, what time to introduce the chip, and for what markets. Moreover, there were complicating restrictions such as that product development had to be synchronized with market windows and that product features needed to meet or exceed the needs of their target markets while satisfying relatively inflexible financial resource budgets.
To solve this problem, the team combined different solution techniques from the standard literature with newer heuristic methods. Their approach was to decompose the problem into multiple stages, use an outer genetic algorithm to search the space, and then subsequently pass information to an inner set of tactical heuristics. They also made a couple of enhancements to the model in order to incorporate strategic goals.
Their new approach has already generated business benefits for Intel in terms of better decision making, increased revenue, and the ability to explore several what-if scenarios. The use of the tool has risen exponentially within the organization, and the feedback from users regarding transparency and ease of use has been favorable.
Congratulations to the 2011 INFORMS Fellows! Fellows are examples of outstanding lifetime achievement in operations research and the management sciences. They have demonstrated exceptional accomplishments and made significant contributions to the advancement of OR/MS.
Kurt M. Anstreicher
Cynthia Barnhart
Oded Berman
David Bell
Izak Duenyas
Eugene A. Feinberg
Harvey J. Greenberg
Avishai Mandelbaum
Robin O. Roundy
Christopher S. Tang
Watch for photo spread in December OR/MS Today and video of the awards proceedings coming to the INFORMS website in December.
INFORMS Expository Writing Award
Ward Whitt, Columbia University - City of New York
Philip McCord Morse Lectureship Award
William Pulleyblank, U.S. Military Academy
INFORMS President's Award
Kenneth R. Chelst, Wayne State University
INFORMS Prize for the Teaching of OR/MS Practice
Jeffrey E. Kline, Naval Postgraduate School
George E. Kimball Medal
Brenda L. Dietrich, IBM T.J. Watson Research Center
Stephen M. Robinson, University of Wisconsin-Madison
John von Neumann Theory Prize
Gérard P. Cornuéjols, Carnegie Mellon University
Frederick W. Lanchester Prize
David Easley and Jon M. Kleinberg, Cornell University
Bonder Scholarship for Applied Operations Research in Health Services
Lauren Cipriano, Stanford University
Bonder Scholarship for Applied Operations Research in Military Applications
James Morris, Air Force Institute of Technology
George Nicholson Student Paper Competition
Kuang Xu, Massachusetts Institute of Technology
George B. Dantzig Dissertation Award
John Turner, University of California - Irvine
INFORMS Undergraduate Operations Research Prize
Matthew Robinson, United States Military Academy
Doing Good With Good O.R.
Turgay Ayer, Georgia Institute of Technology
