I had a great time listening and speaking at the Toronto Data Mining Forum at SAS this morning.
I think that there are some pretty exciting opportunities for web analytics folk to get together with data mining folk over the next three of four years.
Three key points I wanted to express:
-Businesses increasingly want web data combined with business intelligence
-We’re going to have to get along (Las Vegas eloping or shot-gun wedding, it’s our choice)
-We need to start getting to know one other now, (because the differences are quite substantial)
Three key points I took away from the forum:
-The perils of significance and Type I error in segmentation abound. Watch your confidence intervals folks!
-The BUMP algorithm, while greedy, is a really elegant method of identifying clusters of opportunity that are not necessarily linearly distributed, but also goes after those all important normally distributed clusters.
-Ask why you’re modeling something, and go beyond.
It was pretty good, and we’ll see what the Toronto Web Analytics community has to say next week.