Date:         Sun, 12 Apr 2009 13:31:54 -0700
Reply-To:     keqiang.hou@GMAIL.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         keqiang.hou@GMAIL.COM
Organization: http://groups.google.com
Subject:      Low Event Rate Predictive Modeling
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

How is it possible to build a predictive model (either logistic, or
decision tree, or neural network) with more than millions of good
events while only about 300 bad events.
It sounds ridiculous, but we have to find a way.

Anyone with experience in predictive modeling can answer my question?


Many thanks.