Predictive Analytics Didn't See Egypt's Turmoil Coming

Predictive analytics didn't see Egypt's turmoil coming

Predictive analytics software is being touted as a must-have technology for any enterprise that collects data and wants to remain competitive, but do we know how effective it is? When it came to forecasting the political upheaval in Egypt, predictive analytics used by the Pentagon missed the mark, according to a report by Noah Shachtman at Wired.

U.S. defense and intelligence offices have bought more than $125 million worth of computer models over the past three years to predict political turmoil, Shachtman reports. If the investment is paying off, it isn't transparent.

"All of our models are bad, some are less bad than others," said Mark Abdollahian, an executive with Sentia Group, which has sold dozens of predictive models to the government. "But think of this like Las Vegas. In blackjack, if you can do four percent better than the average, you're making real money."

The Defense Department has granted $90 million to more than 50 research operations to come up with "a more predictable prediction system," Shachtman writes, but nobody is anticipating a clear picture into the future for some time.  

For decades the Pentagon has been channeling funds into prediction systems, including Darpa's $38 million Integrated Crisis Early Warning System. ICEWS combines three methods of predictive analysis: Trying to replicate the behavior of key players; examining macro drivers, such as economic, social and demographic forces; and parsing news reports. ICEWS modelers were able to forecast four of 16 rebellions, political upheavals and episodes of ethnic violence.