Date:         Thu, 6 Aug 1998 11:39:16 -0400
Reply-To:     RHOADSM1 <RHOADSM1@WESTAT.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         RHOADSM1 <RHOADSM1@WESTAT.COM>
Subject:      Re: Data categorization for fuzzy match?
Comments: To: "Self; Karsten" <kself@VISA.COM>
Content-Type: text/plain; charset=US-ASCII

Karsten,
 
     Would an off-the-shelf matching package be an option?  This is a
common enough problem that canned packages do exist, and I expect that
they address your issues.  I have heard good things about AutoMatch --
more info is available from www.matchware.com.
 
          Mike Rhoads
          Westat
          RhoadsM1@Westat.com
 
 
<<original posting>>
I am comparing methods of matching membership data based on several key
and demographic fields in very large datasets (100m + records).
 
I need to find ways of restricting the number of potential matches.  I
am looking for ideas or references to:
 
 - Hash or key numeric fields such that transposes and near-misses are
keyed with identical or similar values.  Should be suitable for SSN.
 
 - Hash or key text fields so that they may be searched readily for
similar words and/or text elements.  Should be suitable for name and
address data.
 
Thanks.
--
Karsten M. Self (kself@visa.com)
Trilogy Consulting
 
    What part of "Gestalt" don't you understand?