Those who can, do; those who can't, get certified.
Regular AllAnalytics.com readers might recognize this phrasing from our current Point/Counterpoint debate blog on the value of analytics-related certification. Scott Larsen, an independent consultant, exhorted readers, "Go and do something valuable instead of studying for a certification exam." Harkening back to his days as a data analyst at Google, he explained:
When I participated in hiring committees at Google, lots of certifications was generally considered a negative signal. Usually this came from a feeling a mal-prioritized time -- is there nothing better the job candidate could have done with his or her time? Why not accomplish something? You learn so much more by actually getting dirty doing things than you do studying for a test -- show us where you got dirty and what you learned and what you contributed.
Larsen's advice smacked me upside the head as I read about two data-mining competitions GE recently launched on Kaggle. Could participation in such competitions end, or at least diminish, reliance on certifications as a measure of knowledge? I like this idea -- a lot.
As we've previously explained, Kaggle is a data science marketplace that brings together companies or organizations with business challenges and folks with the desire to tackle them. These challenges are always about bringing out-of-the-box thinking to bear, whether to solve society's thorniest issues, address major industry gotchas, or just have a bit of fun with numbers.
In one new challenge, for example, GE aims sky high -- literally. In tandem with Alaska Airlines, it launched the Flight Quest challenge to address what it says is a $22 billion-a-year problem airlines face in managing efficiency. As a GE Aviation director explains in the video below, the goal is to develop an algorithm that delivers real-time flight profiles pilots can use for en-route decision-making. When pilots have such insight at their fingertips, they can make flights more efficient and reliably on time, or at least that's the stated intent.
The second of GE's latest quests deals with a more down-to-earth concern: healthcare. In its Health Quest, GE is working in partnership with Ochsner Health System to "promote an improved health care system experience for patient and family." But this challenge is about operational improvement, not medical care. The aim is to figure out ways to reduce the "$100 billion wasted annually in healthcare inefficiencies, distracting facilities from their primary focus -- patient care," GE said on the challenge site.
These are but two of many examples of the data-mining competitions going on right now on Kaggle, not to mention other venues. I call them out for their newness -- GE launched each within the last week -- and not because there's anything especially compelling about putting your mind to work in solving flight or healthcare inefficiencies. Neither is a bad goal, to be sure, but my point is that either could provide a great showcase for your talent. Even if you don't win a competition, being able to play around with the big-data sets available to contestants could be well worth the effort.
Next time you're tempted to sign up for a certification class, perhaps you ought to first take a gander at Kaggle. It'll make a great addition to your résumé -- and, who knows, you just might end up with some prize money, too.
Do you have any experience with data-mining competitions, of any size or scope? Share below.
Kq writes Such competitions surely can't hurt. Of course, in reality they're a clever marketing avenue for the sponsor. Geting the company name out there for free in press releases is great advertising.
I'd expect that just about anything a big company does these days has marketing in mind, at least partially, to help justify the expense of the public effort. However, given the examples of competitions sponsored by SAS and others mentioned in this thread, I'd presume that the sponsors expect some kind of valuable output from the competition itself.
... Which leads to my next qustion: I'd wonder if there are examples of actual analytics products now deployed, addressing real-world data challenges, that have been developed through these competitions.
Such competitions surely can't hurt. Of course, in reality they're a clever marketing avenue for the sponsor. Geting the company name out there for free in press releases is great advertising. Whether competitions trump certifications is another matter. Both should be advantageous it would seem.
That's a good question, I don't know. The Netflix and Heritage Health contests were corporate-sponsored, but open to all interested parties, academic, corporate, or (moonlighting?) individuals. Because of their length, those two marathons probably did not get many "student teams" participating as a formal part of their classwork, as some of the shorter duration contests do.
@Doug_Dame, your point is well taken. Do you think it'd be fair to call it the largest data mining competition in academia (vs. the corporate world)? I don't know the answer myself, but think there's a distinction worth nothing here.
Although prestigous and senior in tenure, I don't think it's accurate to refer to the Data Mining Cup as "the world's largest data mining competition."
The Netflix competition lasted almost 3 years, at the 8-month mark had 20,000 teams registered of which 2,000+ had made entries, and paid out prizes in excess of $1 million US.
The ongoing Heritage Health Prize, hosted on Kaggle, is a 2 year contest with a max payout of $3,230,000 and a minimum of $730,000 if the grand target is not achieved. It has more than 1400 teams registered and eligible to compete in the current final segment.
@Beth Kaggle does feature most interesting competitions to get smart people involved in solving data problems. I discovered that NASA does it, too. In October it announced the Launch Big Data Challenge Series for U.S. Government Agencies:
The Big Data Challenge series will apply the process of open innovation to conceptualizing new and novel approaches to using "big data" information sets from various U.S. government agencies. This data comes from the fields of health, energy and Earth science. Competitors will be tasked with imagining analytical techniques and software tools that use big data from discrete government information domains. They will need to describe how the data may be shared as universal, cross-agency solutions that transcend the limitations of individual agencies.