Powerful new data analysis tools can spot suspicious patterns and root out fraud in everything from defense contracts to student loan applications, a collection of federal auditors and technologists said Tuesday.
Effectively using those tools, though, will require smart humans not just smart computers, panelists said at a conference on government data analysis sponsored by SAS, a data analysis and software vendor.
Edward Slevin is director of the computer-assisted assessment techniques division of the Education Department’s Inspector General’s Office. He learned that spotting data patterns isn’t much good if you don’t know exactly what you’re looking for.
“When data mining first came out a number of years ago, it came out from [the Defense Department IG’s office],” Slevin said. “It was the greatest thing since sliced bread basically because DoD had a lot of low-hanging fruit. They had a lot of fraud.”
Other IG’s offices, however, soon discovered they needed to watch for false positives, he said. For example, one person using another person’s credit card might indicate just a sloppy business practice rather than outright fraud.
Even when agencies’ information is as good as it can be, they still need a clear way to translate it to investigators in the field, Bryan Jones, director of the Postal Services IG’s data mining group, learned.
One of the first big data analysis projects in Jones’ office was rooting out fraudulent employee health care compensation claims.
“We had investigators who worked that crime area on our team saying our output looked very good,” he said. “Where we fell a little bit short,” he added, “is that our output happened to be in the form of a spreadsheet, and when you email spreadsheets to investigators you’re not going to get the reaction you expect. Some of them really took the time to understand it and others didn’t.”
Jones and his team turned the spreadsheets into a Web tool with large red dots across a United States map connoting where their analysis showed fraud was likely occurring, he said. Investigators quickly grasped the map’s value, Jones said.
“It’s great if we can understand it,” he said, “but until our stakeholders and our users really understand it and start to use it, we didn’t feel like we were really getting any traction at all.”