M. Eng. Project Slate Builds on Past Successes and Adds New Clients
Each year, Master of Engineering students embark on an activity which is a centerpiece of their professional Masters program, a team-based project to solve a real problem for a client organization. Most Financial Engineering projects are carried out in the fall semester, which students in that concentration spend at Cornell Financial Engineering Manhattan (CFEM) as the third semester in their program. Projects in the other concentrations and minors are carried out throughout the academic year in Ithaca. Read on for descriptions of all of the ORIE M.Eng. projects in the 2009-2010 academic year.
|Lokesh Manohar shows Prof. Henderson, Tim Carnes, and teammates Ben Cheron, CS student Anchal Dube and Kevin Yu a graph of Ornge project data.
More than half of the 2009-10 ORIE Master of Engineering projects in Ithaca are for clients with whom relationships were established in prior years. These projects focus on continuing and refining the work of past teams. New project clients this year include Xerox, Microsoft and the Royal Bank of Canada, while the ongoing clients are Canadian Rail, MITRE, Ornge, Third Battalion, and the Cayuga Medical Center.
ORIE has seen an increase focus on healthcare, including emergency services and disaster preparedness, in recent years. Three of this year's projects reflect that focus:
- A team advised by Professor Shane Henderson, Professor David Shmoys, and Ph.D. student Tim Carnes is building on last year's project for the Toronto-based air ambulance service Ornge to improve the results. Ornge uses 40 aircraft to transport patients throughout the province of Ontario with most of the demand, involving medical transport from one facility to another, known ahead of time. The team's objective is to enable Ornge to run an optimization program every night that generates schedules for the service's aircraft on the following day. This requires detailed understanding of usage patterns, travel times, and fuel costs, as well as data management and computational techniques.
|Professor Friedman meets with Valerie Wilson, Esther Chiew, Aksara Sumethkul and Jonathan Timura about the Third Battalion project.
- A team guided by Professor Eric Friedman is also basing its work on a project completed last year, for a Rochester, New York emergency services organization called The Third Battalion. The organization is seeking to meet a goal of getting their vehicles to emergency sites within 9 minutes, 90% of the time. This entails close scrutiny of usage patterns so that vehicles can be pre-positioned, assigned and backfilled in an optimal way.
- A team under the guidance of Professor Peter Frazier is analyzing the laboratory inventory system at Cayuga Medical Center in Ithaca. The team hopes to identify cost savings from reducing capital tied up in inventory, waste due to expiration, and shipping costs. They are analyzing data on demand and ordering patterns for more than 50 reagents used in laboratory tests to determine the levels that should be stocked to be 99% sure that supplies do not run out.
|Xerox project client Andy Huber points to a particular aspect of the preliminary data presented to him in Rochester, NY by team members Kathryn Bartlett, Cindy Yang, Abinaya Selvam and Paul Delgado.
Xerox is not an entirely new M.Eng. project client, since several projects were carried out for the company in the 1980's. Since then, Xerox has added many services and systems to its product line but continues to deploy equipment, particularly as a means of supplying document services to many companies, and maintains a significant presence in Rochester. Equipment-based product lines require technical services, i.e. maintenance and repairs, that must fulfill contractual provisions relating to response times. These technical services are delivered through customer service engineers comprising nearly 10% of the workforce, plus part time employees and partner companies.
This year's Xerox M.Eng. project team is working under the guidance of Professor Jack Muckstadt to forecast the demand for service, based on the mix, reliability, and usage rates of equipment as well as the obligations spelled out in service contracts. The team will then use this analysis to determine the amount and composition of workforce needed to meet the demand.
| Team members review a draft with Microsoft via videoconferencing. From left, Kevin Liao, Weston Skeans, Xiaocong Wang, Prof. Rusmevichientong, Abishek Ravi.
In providing software to organizations around the world, Microsoft relies extensively on its capability to download programs and updates over the network. Microsoft tracks the characteristics of the downloads as well as customer satisfaction and complaints. The team, advised by Professor Paat Rusmevichientong, is analyzing a large quantity of data to determine the relationship between download success, download attributes (available and required bandwidth, success or failure, etc.), and customer feedback.
|Nate Maichel, systems engineering students Weija Bi and Youyu Zhang, Maksym Samadov and Bo Zhang meet with advisor Peter Jackson and, via teleconference, with Canadian Rail clients in Montreal.
For the past two years, M.Eng. projects for Canadian Rail have received recognition in the form of an IEEE publication and a Cornell Research Conference presentation. These earlier projects dealt with scheduling of rail crews to match employee preferences and the efficient scheduling of crews and trains. This year's project team, advised by Professor Peter Jackson, has the goal of creating a software tool that railroad dispatchers and managers can use to improve the match of supply with demand for crews. The tool is intended to provide statistical forecasts of supply and demand based on data about past assignments.
MITRE, a not-for-profit consulting organization in McLean, Virginia, has engaged a series of M.Eng. project teams to consider specific problems associated with managing and providing essential resources for a population as it evacuates a large urban area. Beginning with work in 2007-2008 and continuing into the following academic year, teams have selected software, established a basic traffic simulation model, and added an optimization layer, using the Washington, D.C. area as a specific example. In the process they identified limits to the rate that people can evacuate the area, and bottlenecks at which queues would develop if the rate is exceeded.
This year's MITRE team, with the guidance of Professor Mark Lewis, is investigating the effectiveness of setting up depots, where the flow of evacuees can be consolidated onto buses. Such an approach raises new issues, such as parking for the vehicles used by those who transfer to buses, determining where best to locate depots, and compliance with the system. In order to deal with these issues at a suitable level of detail, the team will concentrate on one specific route, the Interstate 66 corridor.
|David Matteson meets with Akhil Gupta, Evan Sprenger, Xin Zheng and Irene Shen, members of the team working on a project for the Royal Bank of Canada
Although most financially-oriented projects are completed in Manhattan in the fall semester as part of the Financial Engineering concentration (see below), one year-long project on the Ithaca slate this year is devoted a financial topic. A team is working to predict the outcome of securities trades that pair the purchasing of one stock with selling short of another stock in the same industry, such as pharmaceuticals. The objective of the pairing is to mitigate the risk (variability) associated with the overall market and with the specific industry, with the hope that the difference in value between the paired stocks will revert to the mean. A trader at the Royal Bank of Canada is the client for this project, which is advised by Professor David Matteson.
The Financial Engineering projects that were completed at CFEM in Fall 2009 had clients ranging from Goldman Sachs, Bloomberg, Deutsche Bank and Barclays to the Cornell Investment Office. Two of the projects dealt with portfolio management, two with options hedging, and two with trading strategies.
The optimal allocation of portfolio assets has been a subject of investigation in Operations Research since the pioneering work in the 1950's of Harry Markowitz, who was awarded the Nobel Memorial Prize in Economic Sciences in 1990. His theory and subsequent developments make it possible, using optimization techniques and data derived using applied probability and statistics to determine the mix of assets with the highest expected level of return at a given level of risk, as well as the mix of assets with the lowest level of risk at a given level of expected return. Doing so entails estimating the returns and risks of classes of assets, as well as correlations among the different classes.
- Tax-Efficient Portfolio Management for Goldman Sachs. Often Markowitz-type models do not take tax consequences into account. Assets in a portfolio entail different tax obligations on the money they return to individual investors. In particular, selling an asset for more than it cost incurs a tax at a different rate if it has been held for a year or more than if it is sold before that. Taking this into account can yield better after-tax outcomes, particularly through a strategy known as 'tax-harvesting,' in which short term losses (which offset gains for tax purposes) are taken aggressively and gains are postponed as long as possible. However recent research has shown that there are advantages to taking some long term gains early, in a strategy called 'gain management.'
In one of two projects for Goldman Sachs (see below for the other one), a Financial Engineering team under the guidance of Professor Adrian Lewis used three different Operations Research techniques - Markov chains, Monte Carlo simulation and non-linear optimization, to find tax-efficient portfolios.
- Portfolio Optimization for Cornell Office of University Investments. Johnson Graduate School of Management Professor Robert Jarrow, a member of the Field of Operations Research, has pointed out that the Markowitz-based portfolio optimization approach used to manage Cornell's endowment can benefit from regarding of alumni contributions going forward as an asset to the university that may be correlated with returns from other asset classes in the portfolio. Professor Jarrow advised an M.Eng. project that estimatse the present value of these future contributions, their returns, and the correlations with the remaining asset classes. Working with Cornell's Office of University Investments, they then used optimization techniques to determine an allocation among the remaining asset classes to achieve Cornell's target return with minimum risk.
Among the businesses that banks conduct is the creation and sale of financial products such as options to buy or sell a security at a later date for a specific price. Such options can be used by the bank's customers to reduce their financial risk. For the bank, determining the price at which to profitably sell an option uses formulas (one example is the Black-Scholes formula) that incorporate data about various attributes (parameters) of the relationship between the underlying security and the option. The formulas use past data to estimate the value of the attributes. When selling the option the bank also executes various transactions consistent with the formulas to minimize, or hedge, its own risk. However in times of financial crisis, such as has been experienced recently and at various points in the past, the attributes take on unusually large values, rendering the hedging inadequate and exposing the bank to significant loss and even bankruptcy.
- Options Hedging for Deutsche Bank. An M.Eng. project carried out for Deutsche Bank and advised by Professor Philip Protter undertook to develop and evaluate ways to revise the bank's hedging strategy for interest rate derivatives in response to signals that a financial crisis in brewing. They recommended specific ways to alter the hedging approach when changes are observed in one of the attributes of the relationship between the underlying security and the option. The team noted that such an alteration may reduce profit, but can help insure against intolerable losses.
- Options Hedging For Bloomberg L.P. An M.Eng. project advised by Dr. Sasha Stoikov, CFEM's Head of Research, investigated an approach to hedging of derivatives called exotic options. The approach is based on the idea of replicating such a derivative with a collection of other securities constructed such that the collection has a payoff structure that will always (under suitable assumptions) be higher than that of the original derivative, but without relying on the specific model used to evaluate the original derivative. The team used linear programming to determine such a replicating collection with minimum cost, which could then be used to hedge the original derivative sold by the bank.
Two of the Financial Engineering projects investigated statistical techniques that could be used to increase profits or reduce costs of securities trades.
- Statistical Arbitrage for Barclays P.L.C. Professor David Ruppert and Dr. Stoikov advised a project team that investigated the possibility of taking advantage of the tendency for mispricing of securities to correct over time, so that portfolios characterized by higher or lower than actual value tend to revert to their actual value. The team used trade data to develop several strategies, and tested the strategies against data that was not used in the analysis. One of the strategies, which entailed buying and selling within the trading day, proved promising, even when transaction costs were taken into account.
- Mutual Fund Replication for Goldman Sachs. It has been shown that a large number of actively managed mutual funds do not meet the benchmarks by which they are measured (for example beating the S&P 500 index by a prescribed percentage) despite the higher fees they command. Passively managed mutual funds, such as Exchange Traded Funds (ETFs) are intended to conform closely to the relevant benchmarks and charge less for management. A team advised by David Matteson used forecasting methods to determine which actively managed funds are likely to outperform their benchmarks and constructed a synthetic mutual fund combining a mix of passive ETFs predicted to closely match the performance of the promising active funds.