Posted By:Posted Date:September 01, 2010Points:0Category :Sql Server

I am wondering, if I see a decent lift in a model, but the prediction probability is very low, like 5%, isn't this telling me that the model is a poor predictor of my output? Does this not say that I have a 95% chance of being wrong even though I have
some lift?
Thank you for the help.

Hi, I am using sequence clustering algorithm and I have checked Node probability distribution (state transition) for one of the cluster. Probability values don't sum to 1. Why? I think they should. Do I miss something?

Hi all,
I'm new at datamining, and am wondering about the following problem :
Say I have information about the products a customer buys.
One of the products is milk, and I want to predict milk buyers.
Do you than have to add a variable MilkBuyer with 1 for milk buyers eand 0 for the rest (like the BikeBuyer variable in the AdventureWorks samples) to build a model to predict milk buyers (for instance with a Decision Tree model), or is there a better way?

(newbie)
I have a Bayesian model almost completely overlapping my ideal model in the Lift chart. It indicates 100% of the target population captured but with 0% prediction probability. Basically it rises quickly and that plateau's at 100% following the ideal
model. Do I have something wrong in the problem setup?
If I understand this correctly, this is either a major error on my part or the model completely describes my target population with almost complete likelihood of being wrong?
Thoughts?

## Node probability distibution - values don't sum to 1?