Thursday, November 10, 2011

Predictive data modeling using Kaggle

Kaggle is a startup tech site that lets organizations post datasets and ask world wide users to predict how their customers, clients or patients will behave. It's set up as a competition, and the organizations pay for the best solution with a prize. Here's Kaggle's video explanation:



Kaggle claims on its website always to have outperformed pre-existing accuracy benchmarks, because it draws on a wide range of competitors who keep trying to better each other's work. I can see many applications in social services and yes, you can post scrubbed, anonymous data; you can even post variables as A, B, C without stating what the variables are.

Competitors have already improved NASA's algorithms for mapping dark matter; there's a competition under way to predict hospital stays; and there are classroom applications available so that students can use unknown datasets.

I learned about Kaggle from an Atlantic posting called "The 20 Most Innovative Startups in Tech" by Rebecca Rosen. Check out the article for more interesting websites.

No comments:

Popular Posts