Michael Radzicki has been exploring the use of predictive analytic tools, which mine big data sets for patterns and trends that can help make predictions about future activities or behaviors, in areas as diverse as currency trading and medical fraud detection.
More recently, in partnership with Hossein Hakim,
professor of electrical and computer engineering and an active currency trader, Radzicki has been developing an automated trading system. Using historical data on the Euro-U.S. dollar currency pair and other mathematical indicators, his machine-learning algorithm tests different values to find the combination that would have yielded the most profitable trades. The system essentially mines the data to find and exploit patterns that are so complex and so deeply woven in that humans cannot perceive them. Radzicki has taught classes on trading system develop- ment and has had former students recruited by Goldman Sachs and Putnam Investments. He hopes leading financial services firms will increasingly recruit WPI students, particularly with the launch of the new trading system development program that he and Hakim have established.
Radzicki is also working with a Woburn-based start-
up to build predictive analytics into two software systems designed to link buyers' and sellers' payment systems to make transactions more efficient and transparent. In one, he is using predictive analytics to build an instant credit- scoring tool for companies wanting to join the system and access credit to pay their vendors. For the second, he is using predictive analytics to help detect fraud in health insurance payment data. An early version of his algorithm was able to identify cases of fraud in tests with a portion of the state's data set. He says the state could use this algorithm to pre- vent fraudulent claims from being paid. "Fraud is a huge and growing problem in healthcare," Radzicki says. "By helping reduce fraud, the system can potentially play an important role in lowering costs in the healthcare space." Q
[27]