Date: Thu, 25 Aug 2005 09:23:22 -0400
Reply-To: tmuir@RDG.BOEHRINGER-INGELHEIM.COM
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Tim Muir <tmuir@RDG.BOEHRINGER-INGELHEIM.COM>
Subject: Re: SAS Enterprise Miner - making predictions,
defined model and export model code
Content-Type: text/plain
Hi,
Enterprise Miner generates the complete set or scoring code. After you
build a model, you need to drag over/connect/add the scoring node,
connecting it to your final model, which will produce the SAS code.
-Tim
-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU]
Sent: Wednesday, August 24, 2005 5:36 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: SAS Enterprise Miner - making predictions, defined model and export
model code
Hello,
I have some questions regarding the toolSAS enterprise Miner.
(For anyone intrested,here is the tutorial that I followed
http://www-staff.it.uts.edu.au/~mbeaureg/topics/prediction_in_sas/prediction
-sas-tut.pdf )
I have my model and a production dataset connected to a score node. this
node is connected to an Insight node.
My production set has 70,000 records. However, after finishing run the
insight node, it showed me a table of 80,000 records. I dont know if this is
the predicted values for the score data node or not (more than the
production/score data).
So I tried another way. I created a production dataset by taking a dataset
(with the output coulmn). I make all of the values in that column to "."
(missing value). Add a data node with that column as role of PREDICT. The
same result happened (output 80,000 while my production data has 70,000
records.)
+ Could you please give me some hints on this problem? (I dont know am I
doing rite for making predictions)
+ There are some models that are not built in SAS, such as Gausian
Classifiers, or Genetic programming. So if I want to implement these
algorithms, what shoudl I do? I saw a USER DEFINED MODEL node but not sure
if this is a clue.
+ After building a Neural Network in SAS Enterprize Miner, I would like to
write some manual code (in the programediter) to call the built model. Does
SAS provide any EXPORT function to save my Neural Network (weights) for
latter use? I saw a node called SAS CODE but I am not sure if this is a
clue.
Thanks
Have a nice day
Patrick Tran
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com