Peerside

From OSNews Comics

As most modern data wranglers know - big data and machine learning are the new black and on everyone’s future career path whether they like it or not. The folks at UC San Diago along with FICO are running a data mining contest where the challenge is to maximise accuracy of binary classification of a provided ecommerce data set. The prizes are nothing like the famed Netflix treasure and it’s limited to college students but even then, like the contest flyer points out, the best way to really learn machine learning is with practical experience – learn by doing.

So, if you’re interested read the contest flyer below and check out the 2009 UCSD Data Mining Contest home page

Welcome to the 2009 UCSD Data Mining Contest sponsored by FICO.

FICO has sponsored the data mining contest since 2004, which provides undergraduate and graduate students an opportunity to test out their data mining skills on a real-world data set. FICO pioneered the use of predictive modelling to represent and explain the underlying relationships in data and make predictions and classifications about future events.

This year's contest consists of two classification tasks based on e-commerce transaction anomaly data. The first task is to maximize accuracy of binary classification on a test data set, given a fully labelled training data set. The performance metric is the lift at 20% review rate. The second task is similar to task 1, but provides a couple of additional fields that have potential predictive information.

FICO and UCSD will award prizes to first, second and third place winners in four categories: Task 1 undergraduate; Task 1 graduate; Task 2 undergraduate and Task 2 graduate. There is a total of $4,000 in prize money for each task, for a total of $8,000.

We launched the contest on May 15, 2009. The contest will end July 15, 2009. The contest is international. All current undergraduate students, graduate students, and postdoctoral researchers studying full-time, in residence at an accredited university or college may compete for prizes. Others may compete but will not be eligible for prizes.

---

• The UCSD Data Mining Contest offers students a chance to test their data mining skill on a real-world data set.
• Teams will compete to build a system that correctly classifies data from an e-commerce website.
  See the datasets page for details.
• There are many reasons to participate:
  1. Practical experience is the best way to learn.
  2. Success in the competition looks great on a resume (graduate admissions, jobs, fellowships).
• Participating is easy, and we encourage everyone to try!
• The flyer for this competition.
• The contest forum is now up and running. Learn from others and get your own ideas out there!

Recently I've been focused very much on work related projects and some very cool stuff I'm hoping works out next year. Coming up to Christmas I'm hoping to turn my work brain off, clean up my online identity, reduce more paper from my life and get back to blogging - with Wordpress. I had considered spending the time to pick up the most recent Drupal release and updating this blog but recent experienes of other bloggers upgrading with module incompatabilities and a nagging attraction to both Buddypress and the new Wordpress UI convinced me to take the leap. Hopefully it's worth it and I'll end up with a cleaner blog that I can maintain with less effort. For anyone not up to date with what Wordpress are doing check out the recent Coltrane release

One of the lessons I've learned over the years is that nothing creates a happy customer more than helping them solve their exact problem - after all that's why they bought your software, right?  In many cases this never happens even when your product is the solution to their problems because modern software products are developed, marketed and documented in a generic "one size fits all" model which, while attempting to be as generic as possible, documents individual features but not end to end use cases. The enterprise software market, where software suites are often large and complex, this phenomenon has made the existence of expensive consultants or paid support a fact of life.

Yesterday, while evaluating PBWiki for a personal hobby site, I signed up, created, evaluated and then deleted an online Wiki site within the space of 20 minutes ( yes - 20, imagine the same for Sharepoint). The process was everything it should be, simple, obvious and direct, but what struck me as most interesting was an email I received from the PBWiki team immediately after signing up to offer me free access to a weekly customer support session - no strings attached.

One of my weaknesses is that I hate to dump books, sell them - fine, share them - ok but I just can't bring myself to dump a story. I'm also the type of person who can sink into a book when travelling and as a result I often find myself away from home with one or two finished books and trying to decide what to do with them. I used to leave the books in the hotel room hoping that the cleaners may find them useful but that was until recently when I saw a pile of books dumped in a skip outside a hotel in Dublin - horror !!!

So recently when I spotted mollydot releasing books via bookcrossing.com I decided that in the future I too would give my books freedom and release them into the wild. This morning I released my first book which can be viewed here , set free in the Sofitel lobby in Redwood City. Here's hoping someone else gets to enjoy it .. and if they dump it then let it be on their conscious, not mine !

Bookcrossing, for those of you unfamiliar with the activity, is a way to tag and share books through an online community Bookcrossing.com. As the web site describes itself:

Bookcrossing is away to share your books, clear your shelves, and conserve precious resources at the same time. Through our own unique method of recycling reads, BookCrossers give life to books. BookCrossing books are not stagnant dust collectors, but living entities travelling the world as true BookCrossing emmisaries. Our books find new readers and introduce them to the wonders of BookCrossing.

I'll blog some more on bookcrossing at some stage in the future, I find the idea intriguing and it intersects with an idea that has been nurturing in my subconscious for a number of years now about how we can efficiently combine book recycling with libraries, charities and social communities. More on that later. In the meantime - get out there and set those books free !

Blogging, tweeting, your linkedin feed, facebook, flickr gallery, GPS location, who you're talking to on the phone, what you're wearing, how you're feeling ... where will it all stop, or will it ?

I'm currently reading Accelerando (Singularity) by Charles Stross which starts with a character who makes his living from spreading ideas, hooking people up with new technologies, memes and innovations - interesting stuff but loses the plot about half way through. What I found intriguing was how, in the near fictional future, not only does everyone lifestream everything we see, hear and experience but they combine this data with personal agents for research, background checks, correlations etc. Imagine having your social agent whisper peoples names, when you met them last and the name of their significant other as you meet new people - not because you have told it that information but because it has mined your lifestreamed recordings and inferred the information. Another cool geolocation scenario is suggested where one of the characters ventures into a dangerous part of the city and is reminded by one of her personal agents that this will impact on her insurance premium - if the agent stuff doesn't freak you out then the use of geo data for risk based insurance billing should.

Have fun on twitter ....

After my last post Apple, iPods and Data Quality I've had a few queries from friends about what exactly I do in Informatica :-) Nothing new there but what did surprise me is that this time most of those people were aware of data quality because their CIO, Compliance department or customer services guys were making noise about it and to because they saw it as something they had to think about it in their day to day lives. A few years ago when we were still Similarity Systems and I told people I worked in data quality I may as well have said that I polished widgets ... no one had heard of data quality or worse just assumed it was data cleansing.

I believe that will change for individuals as well as companies as the world we live in becomes more and more driven by data. The case for companies is quite clear - data quality issues lose money - period. For individuals I believe the same case is there but we rarely measure our activities based on cost so it's not as apparent. Think about the personal cost to you for example if your credit check report ( especially stateside ) is incorrect, if your address is incorrect and your tax refund goes astray or your timetables are wrong ...

And on that last one I leave you with the most recent Informatica marketing video - ok so the acting is terrible but the punch line is worth waiting for.

I rarely mention my day job in my blog because mostly I blog about alternative topics to give my brain a break - isn't there a saying "a change is as good as a rest" ? However today is different, I quite innocently stumbled across a classic data quality issue in real life - and all because I bought an iPod and wanted it delivered to my brother in Waterford.

The Apple Store is great, allows you to select what you want, select the color and even engrave a message on the back. No complaints there. So finally, time to enter the destination address, no surprises there and I provided a perfectly correct Name, Townland, Parish Name, County address and off it goes. Apple prompt send me my confirmation and the next day I get my order reference and UPS tracking code.

Twitter, Jaiku, Pownce, Tumblr, Friendfeed and the recently added Identi.ca - choices, choices and more choices. Surely if microblogging is going to stick around it'll have to standardise on a messaging format (XMPP anyone ?) and the tools just become end points just as Outlook, Eudora, GMail, Hotmail etc. became endpoints for email ?

In the meantime I'm trying out Posty to mix 'n match them all.

As much as Apple deserve credit for their recent, market changing, innovation around the iPhone they are by no means the only smart people out there. Take a look at the concepts bubbling out from the Aurora project created by Adaptive Path as part of the Mozilla Labs concept series.

I really like the idea of my UI adapting to my location and deciding what applications I most likely want to use and what data I'm most likely to find interesting. Similarly I like the tactile like gestures shown in one of the earlier videos - there's something about how sliding the menu out slowly makes sense yet a quick, dismissive, flick can exit back to the desktop (or should that be phone top ?).

I think the next few years are going to get very, very interesting as pervasive connectivity and location awareness change the way mainstream society thinks about how they interact with their surroundings. When your location becomes a program parameter does that mean you can customize applications for particular locations ?

Aurora (Part 1) from Adaptive Path

Aurora (Part 2) from Adaptive Path

Aurora (Part 3) from Adaptive Path

Aurora (Part 4) from Adaptive Path