Software
From: Clemens van Brunschot c.vanbrunschot@chello.nl
Date: Sat, 30 Jun 2001 13:51:42 +0200
Subject: SAS data mining macros revisited
In several earlier editions of KDnuggets mention
was made of SAS macros that have been written for speedy development of
predictive models for 0/1target variables. There is news about these
macros.
A new macro was produced together with Matt
Redlon of fingerhut.com. The macro Collin reduces multicollinearity among
predictors by iterative automatic removal of variables according to four
criteria in a linear regression analysis.
- The core transformation macro,
Optibin, now has improved class boundary management, stores information about
transformed variables, and allows the predictor data to be processed in
chunks, speeding up the analysis.
- The macro Interact also stores information about transformed variables.
- The macro M_Spec has received a major overhaul and
now translates a logistic regression model on transformed variables
(including interaction dummies) to SAS code in terms of the original variables,
using the information stored by the Optibin and Interact macros. So that
the model specified can be run on any similar file for an assessment test or for
production.
For more specific information see:
http://members.brabant.chello.nl/~c.vanbrunschot/macros.html
Clemens van Brunschot
| |
|