Managing the Changing IT Landscape: Big Data Analytics
There are multiple approaches to building data analytics models to solve a problem. The Intel-sponsored March Machine Learning Mania Kaggle competition has shown us that. The second stage of the contest inspired 440 entries—that’s 440 versions of a predictive analytics model —from the 251 analytics teams. My congratulations to the winner Grimp Whelken, a Kaggle novice, whose data model best predicted the winners of the NCAA men’s basketball tournament. Here are the final top 10 teams from the competition.
Lessons for Business from the Big Data Dance
Throughout this contest I’ve been interested in how elements of this predictive analytics contest can apply to business and IT. My previous two blogs in this series explored the skill shortage of data scientists and the importance of real time prediction. As the competition closes and I write my final blog, I noticed that the top 10 teams are almost all single players. But were they working alone?
In the competition, Kaggle players were supported by data derived from a sports ratings website and this information was pre-packaged and provided to the contestants by Kaggle.
In the real world, identifying the appropriate data, prepping it for use, and validating the model requires a close partnership between IT and the business. Business analysts with their domain expertise often provide a critical perspective on a big data problem and its relevant data. In the Kaggle contest Ken Massey, someone with recognized domain expertise, essentially served that role for each team.
Small Teams, Big Potential
Intel follows a big data project model that empowers small teams to pursue initiatives that can be accomplished in six months and promise $10million in return on investment (ROI). Intel has successfully completed more than a dozen such projects, and is now pursuing higher value opportunities with $100 million ROI. These five person teams (much like a basketball team) tap a variety of skilled positions including IT, advanced analytics, and business experts.
Intel’s approach is validated by Tom Davenport, author of the recently published Big Data at Work. Davenport found that the large companies he interviewed for his research were forming teams of people with a range of skills rather than hiring PhD level data scientists on a large scale. The teams included people with quantitative, computational, and business expertise as well as skills in technology, and change management.
Teams bring together the range of skills needed to tackle big data projects—business, analytic, and technology. How does your organization partner with the business to deliver projects of high ROI?
Personally, I was very excited to see both UCONN Men and Women win an historic sweep of both NCAA basketball tournaments. Go Huskies! Looking forward to next year to watch the prediction engines in action again!