Good feature extraction is always the major key to response studies like this. With SGD, there may be a bit of additional variable scaling.
On Mon, Apr 25, 2011 at 2:41 AM, Sean Owen <[email protected]> wrote: > I'd be interested to take a run at it with a classifier. Seems like a very > straightforward problem, so, I imagine it's only an interesting approach > that will do especially well. > > On Mon, Apr 25, 2011 at 2:28 AM, Danny Bickson <[email protected] > >wrote: > > > After some discussions of Yahoo! KDD Cup started by Sean Owen on this > > newsgroup, > > I got this note from Metha > > Krishna<http://www.thinkalytics.com/service.php?id=6>, > > co-founder of Thinkalytics: > > * * > > *Hi, > > Check out another datamining competition to keep you busy after the KDD > CUP > > @ www.hearstchallenge.com . $30k in prize money and a lot of fun! > > Regards, > > Krishna* > > > > This is an interesting contest, where the task is to classify user > reaction > > to emails.* * > > Release date for the actual dataset is June first. Submission is Aug. 31. > > I wonder if anyone is planning to use Mahout for this contest? > > > > > > > > - Danny bickson > > >
