March 16, 2007 -- Take the pattern in a Persian carpet: it’s intricate, but totally static. Because the threads in the carpet remain relatively inert, the carpet will never change, no matter how long you stare at it.

In Ginger Davis’ world, patterns evolve as their constituent elements—some visible, some hidden — change over time and interact in a variety of ways. The assistant professor of systems engineering uses multivariate time series analysis to track these changes and understand their significance.

Davis, shown right, applies these tools in a variety of contexts. One area is computer security, where she analyzes patterns in the flow of packet data. She can look at the arrival times and patterns of packet data for applications that typically run on a particular server and apply a hidden Markov model to determine the intangible as well as the observable parameters of the system. She then uses this information to identify the characteristic patterns of data flow associated with permitted applications, creating a basis for screening out other activity. “This system could be used by a company trying to prevent employees from using AOL Instant Messenger or to prevent someone from hacking into a backup server,” she says.

Her work is ideally suited for environmental phenomena, where there are many complex variables that must be tracked and correlated over time. For instance, she is working with government officials in Houston to make real-time predictions of ozone events, evaluating real-time data generated at 42 stations located across the sprawling city.

Davis is also working with the U.S. Geological Survey in Missouri to track the spawning behavior of an endangered species of sturgeon. In the past, the only way to tell if these fish were ready to spawn was to capture and kill them. Davis has taken data from fish that were equipped with sensors recording water temperature, depth and location. “We were able to determine that there were characteristic differences in the depth profiles of fish that spawn and fish that don’t spawn,” Davis says.

Finally, she is exploring leading and lagging relationships among stock markets as the trading day moves around the world. She has developed a technique to look at the return on an asset in the U.S. market and predict the return on similar assets in overseas markets. Based on the performance of an exchange-traded fund such as an S&P 500 Index fund at closing in New York, an investor can buy or short an exchange-traded fund on another market. “You only trade when the forecast of price and variability are favorable,” she remarks. Davis has found that an investment reflecting this strategy would turn a profit over the course of a year.

From U.Va. Engineer, Fall 2006, Volume 19, No. 1

Written by Charlie Feigenoff, a freelance writer who specializes in presenting new technologies to generally educated readers. In addition to his work on University publications, he has written corporate marketing materials and annual reports.