[Skip Global Navigation]

Bienvenue sur le site de SPSS Maghreb
Finding votes one by one
Mercredi, 12 Mars 2008 00:00

Predictive analytics keeps improving. With enough information, the software can anticipate your future wants and needs. NetFlix is a prime example of a business that can help you find movies that you've never heard of, but will really like. And Ken Strasma, president of Strategic Telemetry Inc. of Washington, D.C., uses predictive analytics to locate undecided voters and match them with a politician's position they'll also really like. But Strasma says it's not just better software, it's the hardware, too.

"The real competitive advantage we have is our machine-learning algorithms," he says. That is, he can run his predictive analytics software on clusters of low-cost Windows-based multi-core servers for more precise answers to more complex questions about voter preferences. His distributed network gives him economies of scale so he can handle more political campaigns-112 races in the 2006 election cycle and that many or more for this year's pollfest.

Processing time is critical during a crunch election time as now, Strasma says. So, in addition to hardware, he's adopted Clementine, a predictive modeling tool from Chicago-based SPSS Inc. He says the software runs 20 different algorithms simultaneously and finds the ones that work best on a given data set. "It saves days of processing," he says.

And time is critical with more than 100 campaigns at stake, all seeking to locate pockets of undecided voters. But Strasma is confident with the right combination of data, analytic software and hardware horsepower, he can find every one and even which issues they feel strongly about. Then he can align each voter with one of his clients' positions. For example, he points to evangelical Christians who are mostly conservative, but also ardent environmentalists and will consider voting for Sen. Barrack Obama on that issue alone.

He concludes, "The most costly part of any campaign is getting voter attention." Increasingly, predictive analytics is the tool of choice for politicians who want to get noticed.

Source : Mark Hall - computerworld.com