Dr. Margaret "Maggie" Dunham, Professor, Department of Computer Science and Engineering
To call Maggie Dunham a pioneer is no exaggeration. Dunham began working with computer databases before the word "databases" even existed.
The friendly professor with the warm smile earned Bachelor's and Master's degrees in math from Miami University in Ohio in 1970 and 1972 and then set about to put her talents to use in industry. She started off working for Xerox in upstate New York, doing database types of activities, though the word wasn't in use at the time. She later worked for Texas Instruments and Anderson Clayton Foods, continuing to work with computers and learning all the pitfalls – most caused by human error – that can occur.
Deciding that the world of academia was a better fit for her inquisitive mind, Dunham enrolled in the doctoral program in computer science at SMU, earning her Ph.D. in 1984, and becoming a member of the university's faculty not long after.
Today, Dunham directs the school's Database Research Group. Since 2002, Dunham's major research emphasis has been data mining, which she describes as "taking data and learning something from it that's not obvious."
For a simplistic example, imagine you had a list of individuals and their birth dates. You could take that information and determine their ages from that information. That's not true data mining, Dunham says, but it gives an idea of the type of thinking involved.
For a more realistic example, think about a government agency identifying possible terrorists via information about travel, phone call patterns, banking patterns, etc. "My research is looking at how to study data and come up with new algorithms for prediction," Dunham says. The professor wrote a book on data mining that was published in 2003 and has become a standard, translated into numerous languages.
One of the applications she is currently working on, in conjunction with Dr. Jim Yu, assistant professor of environmental engineering, is predicting the biodegradation of various chemical compounds. There are five or six existing models to predict how easy it is for bacteria to get into a substance and break it down, but while the models each have their strengths, they also contradict one other and each has limitations, Dunham says.
Their project involves looking at various known factors such as freezing point, boiling point and molecular weight for compounds like phenol, glycerol and benzene and trying to tease out correlations with breakdown rates. In most data mining projects, the researchers are presented with a set of data, but one of the challenges of this particular project was that they had to come up with a set of data to work with. "I had to figure out what data was available and what made sense," Dunham says.
The breakdown rate of chemical compounds is just one of innumerable potential applications for data mining techniques, she says.