You will be redirected to the page you want to view in  seconds.

Data analysis helps feds predict future

Jul. 25, 2010 - 06:00AM   |  
By SEAN REILLY   |   Comments
The Housing and Urban Development Department plans to analyze a wide array of risk factors to identify circumstances apt to push individuals into homelessness.
The Housing and Urban Development Department plans to analyze a wide array of risk factors to identify circumstances apt to push individuals into homelessness. (Getty Images)

This is how the Federal Aviation Administration seeks to prevent airline accidents nowadays: connect the dots between dozens of databases encompassing everything from air traffic patterns to weather reports.

It's an example of how federal agencies are trying to make better use of the mountains of information they amass via a high-tech forecasting technique known as "predictive analytics."

Under that program begun two years ago, FAA routinely crunches vast amounts of air travel data in hopes of pinpointing systemwide accident threats. The goal: Deal with the threats so accidents don't happen.

Similarly, the Housing and Urban Development Department plans to analyze a wealth of risk factors to identify circumstances apt to push individuals into homelessness.

The findings will be used to gauge where homelessness is likely to increase and then to direct resources accordingly. The goal is to virtually end homelessness within a decade.

At the Centers for Medicare and Medicaid Services, several predictive analytics demonstration projects are underway to combat improper payments, a spokesman said, with one endeavor employing "automated detection technology" used by companies for fraud prevention.

Although definitions vary, predictive analytics roughly means analyzing data from past events to predict future trends.

"The essence is really using statistical modeling to understand both the past and present and the future of the world around us," said Norman Nie, a pioneer in the field who now heads software startup Revolution Analytics in Palo Alto. Calif.

The technique is not new. What is unprecedented is an exploding trove of digitized information coupled with the computing power to make sense of it.

With its Aviation Safety Information Analysis and Sharing system, FAA fuses 46 safety databases containing millions of records.

Traditionally, the agency relied on the results of accident investigations to guide safety improvements.

With commercial air fatalities at a 75-year low, there aren't that many accidents to investigate any more. But with air travel expected to rise significantly, safety improvements must continue in order to hold accidents down, said Jay Pardee, director of accident investigation and prevention at FAA's safety branch.

Rather than wait for the next disaster, Pardee said, "we wanted to be getting out ahead and leverage the vast quantities of safety information that are available."

After focusing on incidents when planes came too close to the ground, for example, the agency worked with the airline industry to voluntarily install the latest version of a program intended to prevent aircraft from hitting mountains.

The agency's safety system uses both commercial and customized software; the annual price tag of $18 million to $20 million is far less than the $200 million cost of a wide-body commercial jet crash, Pardee noted.

How many other agencies use predictive analytics is unclear. The informal consensus, however, is that the number is growing.

In fields like energy, health management and defense, the level of interest is unprecedented, said Sid Probstein, chief technology officer for Attivio, a Massachusetts maker of analytics software.

Forty-three percent of federal financial executives surveyed recently by the Association of Government Accountants said their organizations make regular use of predictive analytics for everything from staffing needs to investment strategies.

The annual AGA survey asked the predictive analytics question for the first time this year, said lead organizer Clifton Williams, a partner at accounting firm Grant Thornton LLP.

Williams' view is that federal use of predictive analytics is on the upswing, but only in pockets. "It hasn't been linked into the mainstream program processes of what these agencies do," he said.

Vendors are touting the potential.

"It's a phenomenal opportunity," said John Hardigree, the partner responsible for global government analytics at Accenture. In February, the consulting firm announced formation of an "analytics group" with software developer SAS aimed in part at the public sector.

Others strike a more cautious stance.

"There's a lot of hype around it," said Lee Tien, a senior staff attorney with the Electronic Frontier Foundation, a nonprofit organization that works on free speech and privacy issues. "But there's a lot of power around it."

In 2002, the Bush administration triggered an outcry with its Total Information Awareness program, which relied heavily on data mining and some elements of predictive analytics to find purported patterns of terrorist activities.

Amid questions about the program's leadership and its impact of Americans' privacy, Congress shut it down in the next year, although at least some aspects are believed to have survived under other auspices.

Predictive analytics "has enormous promise," Nie said, speaking generally, "and enormous potential for abuse."

More In Agency News

More Headlines