Groups | Blog | Home
all groups > sql server data mining > november 2003 >

sql server data mining : reading analysis server results - decision trees


jason sirota
11/17/2003 10:45:38 AM
Thanks to Peter Kim's detailed instructions, I ran the a
simple artists to customers mine and was successful in
retrieving the data, but I am having trouble reading the
results from Analysis Server.

It shows a number of colored boxes connected by
missing/exists lines. Here is my question:

Let's say we look at the tree for Product A. There are 2
branches for Product A:

Path 1: A exists, B exists, D exists
Path 2: A exists, B missing, D exists

I realize that Path A is the probability that A,B, and D
exist together. I also understand that Path 2 is the
probability that A and D exists but B is missing.

What I want to determine is what is the probability that
A and D exist regardless of whether B exists.

Do I have to add the two together? Is there anyway to
read directly the correlation between A and D?

Thanks!
Jason

Peter Kim [MS]
11/17/2003 2:14:40 PM
Please check out my answer to your original thread.
--
Peter Kim
This posting is provided "AS IS" with no warranties, and confers no rights.

[quoted text, click to view]

ZhaoHui Tang
11/17/2003 2:19:26 PM
Related to your question
" What I want to determine is what is the probability that A and D exist
regardless of whether B exists. "
You may create a simple decision tree with two columns: A and D, both Input
and Predictable. In this way, you can see their correlations.

There is a new algorithm in SQL Yukon which is more specialized in pairwise
correlations: Naive Bayes. You may apply for SQL Yukon Beta 1 sofware. If
you would like to try, please contact me.

regards,

ZhaoHui

[quoted text, click to view]

AddThis Social Bookmark Button