John Elder is one of my favorite data mining professionals and his presentation on top 10 analytic mistakes is by now textbook material. I am mostly amazed at the mistake number 10 – but I must point out that it happens so often in industry, that I consider that to be baseline, and not really a mistake.
I think it is just intuitive for most of the data mining people to try to go with one “winner” model, rather than working on an aggregation approach. Optimized ensemble models is really my specialty, and it is fascinating to me how much resistance sometimes the ensemble models face.
Looking forward to presenting “Inland Transportation Improvement using Trade and Transport Data” at the World Bank tomorrow (Oct 26th, 2010) with my colleague Larry Kahn, a Director and member of the board at NTELX.
For second consecutive day, PAW DC dazzled. First an excellent panel by David Duling, Usama Fayyad and Jason Verlen, and then excellent presentations by Nick Besbeas and Usama Fayyad. Couple of good insights mentioned were the case of belly button rings at Nordstrom, and the positive impact of those annoying banner ads.
Miner3D's Beata and Dusan Toman, and Amrinder Arora
Had a productive day yesterday at the PAW DC. Good time meeting old friends and new people. Some very good presentations – especially John Elder’s presentation on Text Mining, and Karl Rexer’s presentation on industry trends were very well received. Also well received were presentations by Daymond Ling (CIBC) and Dean Abbot. I especially liked Dean’s talk on survey analysis for YMCA, since that didn’t just stop at analysis, but also talked about how to make analysis actionable, and how YMCA branches were able to meet their business goals by taking those actions. Dean also refers to the excellent article “The One Number You Need to Grow” by Frederick Reichheld, which makes plenty of sense in case of YMCA. Also, as in case of other surveys, multi-collinearity is an important factor that had to be dealt with.
John Elder, Gary Miner, Amrinder Arora at PAW DC
Also had a good chance to connect and reconnect with Tom Fuyala (from 11Ants), Michael Zeller (from Zementis), Dusan Toman (from Miner3D), Mark Norton (Idiom) and plenty of other folks.