Time Off

October 2nd, 2008

age moshiach purim topamax and back pain diflucan 150 is generic viagra from india safe aleve and red blood cells coumadin weight loss all uses for topamax breast augmentation expert michigan ansaid medication names buy altace without prescription poly brite bulbs synthroid forums anxious after taking lexapro long-term nexium use coral calcium and fibromyalgia imitrex with nsaids gi problems and arava whipped cream gives you acne soma god prograf and aminoglycosides virginia abana penile burning sensation allergic to diflucan phentermine and lexapro book prozac nation plavix class action suit tramadol tenuate printable light brite patterns rancho allegra lisinopril alchohol prozac nation script tramadol make you tired dose of zoloft cardura medicine identification tramadol prescribing information geodon adolecents irregular menstrual bleeding synthroid itching rash lamictal lamotrigine takes motrin drug manufacturer seroquel for depression levaquin lawsuites relafen vs vioxx vivitrol and cymbalta mixed patient review of aciphex paxil quit hotel robinson club soma bay aricept 60 mg subcu daily clinical trials arimidex richardson labs hoodia gordonii 500 coumadin education prometrium prescriptions

Well, this week things have been pretty crazy. My editor is sick. On top of that, when I tried to do some hardcore predictive modeling on my turn signal dataset, I found out there wasn’t much of an overall predictive pattern in the data, no matter what model I used.

The data modeling I was doing was in a program called Rattle. Rattle is a superb open source front-end for the equally open source statistics program (and programming language) R, which I briefly mentioned in this post. In that post I mention a really great class on data mining, which you can watch in its entirety as a series of Google “Tech Talks”.

I have long since gone through all the material in that class. I used a variety of data modeling techniques I learned (decision trees, boosting, bagging, support vector machines, etc.) but all they did was tell me that my turn signal data probably had a lot of noise and randomness in it (which is good to know, of course). I tried looking at other past data sets, but nothing really worked as well as the turn signal data.

On top of that, the Tampa Bay Rays played in (and won) their first ever playoff appearance today (they finished last in the league last year), the undefeated University of South Florida Bulls face off against Pitt tonight, and of course the Vice Presidential debate is also on. So, in light of these extenuating circumstances, I’m going to cut this post short. In the meantime, I’ll be searching for better data sets to try my modeling on. If you have any interest in predictive modeling, you really owe it to yourself to try Rattle. Combined with R, it’s probably the best and most useful data mining tool that currently exists, plus it’s 100 percent free and open source. And it’s very easy to use once you get the hang of it (and you actually know what those predictive models are doing). There’s really nothing else like it. See you next week!

| | del.icio.us

One Response to “Time Off”

  1. Kenny Says:
    October 2nd, 2008 at 9:30 pm

    Things aren’t looking so good for USF at the half. The Fighting Wan-staches are up for this game, and South Florida keeps having D-linemen drop like flies.

Leave a Reply