Kaggle contests make sense of data

Written by Paper. Posted in Crowdsourcing, New Media

Tagged: , , , ,

Published on February 02, 2016 with No Comments

The latest buzz headline seems to involve the words, “Big” and “Data.” The Presidential campaign will use “Big data.” How to manage “Big data?” How much storage will we need for “Big data?” But if we think of data as just a collection of numbers (or ones and zeroes), it’s clear that data is nothing without the ability to make sense of it and use it to make useful predictions. Enter Kaggle. Kaggle is the place where data meets its bosses. Mathematicians, statisticians, economists and really, really smart people who like to play around with numbers employ all sorts of data sets to make predictions on just about anything. At the Kaggle site there is an eBay auction style list of current competitions where the aforementioned smart people can win prizes from companies and organizations that want their data to make sense.

One of the more interesting past Kaggle competitions involved crunching the numbers on grocery shoppers to predict when they’re likely to visit the grocery store and how much they’re likely to spend. A successful prediction was counted when the date of the shopper’s next visit was accurate and the amount spent was within $10 of the prediction. The winner was a math professor in Russia, who did it with no real world knowledge of U.K, customers, their language, or their cooking, eating and shopping habits. The prize for that was $10,000. A current Kaggle contest with a $3 million prize expires in 10 months. The goal is to analyze the data from members of a health care plan to predict how many days a given patient is likely to spend in the hospital within the next year. An accurate prediction could mean significant cost savings made possible by early intervention and preventative strategies. Currently, more than 1,000 teams are working on this problem.

Share this Article

About Paper

Browse Archived Articles by

No Comments

Comments for Kaggle contests make sense of data are now closed.