Date: Fri, 14 Mar 1997 10:49:49 -0800
Reply-To: dymond@CCNET.COM
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: Anthony Dymond <dymond@CCNET.COM>
Subject: Re: Pre-SUGI Data Mining - KDD Discussion
In-Reply-To: <33273FB5.3608@ccm.fm.intel.com>
Content-Type: text/plain
On Wed, 12 Mar 1997, Thomas Skinner <Thomas_M_Skinner@ccm.fm.intel.com> wrote:
>Dear SAS-Lers,
>
>In an attempt to provoke a bit of discussion on the topic of Data Mining
>a.k.a. Knowledge Discovery in Large Databases (KDD), I thought I'd start
>a thread, pre-SUGI.
>
>Here are a few of the questions I've come up with:
>
Tom,
Lots of interesting issues -- here are a couple of notions:
>How comfortable would your organization/company be with a "Black Box" /
>Proprietary solution?
I don*t think a "black Box* solution is feasible. A variety of
different tools are needed for each situation. Human interaction is
also of great importance.
>
>Do you perceive this discipline as a "Ph.D. Required" exercise?
Not really. You do need to know enough to learn what you need to know
next. (I wore out the door hinges at the UC Berkeley math/stat library
on my last project.) A data miner needs to produce practical pragmatic
solutions to novel problems. People and management skills are at least
as important in order to lead/interact with the local subject matter
experts.
>
>Do you expect to utilize SAS for any/limited/all aspects of this
>exercise?
Yes, extensively, continuously, (almost) exclusively. Nice toolbox!
>
>Do you expect a consulting/services firm to provide the expertise to
>start/administer such a project? Or do you prefer to embark on this
>endeavor independently?
(I have a kid starting graduate school, so I certainly hope consultants
are going to do the bullk of data mining!)
I suspect that data mining may be outside the scope of many smaller in-
house SAS teams. Data mining can be esoteric even for a very good SAS
team. One group can*t do everything.
>
>If you are a consultant, do you plan to add this service to your
>repertoire?
Yes. Data mining can really leverage the returns from a data
warehousing project.
>
>Do you see this activity as a macro or micro-mining activity? Based on
>hardware? Database Size? Platform/Tool Capability? Other?
By the time the subject matter experts are set up and preliminary data
warehouse activities are completed, this can be a fairly significant
project. However, I can imagine a small project where everything is
already set up and the mining questions are fairly narrow.
>
>Are agents/rovers in your future plans?
I did an expert system shell a few years back (in MUMPS) and have been
thinking about trying to rebuild it in SAS with a Knowledge Base to
impliment a data miner. (I lack only the 300 man-years of spare time to
do this.)
>
>Is Visualization commonly used in your organization? Is it already
>applied in a Data Mining context? If so, for interpretation of
>results? Model development? Other?
Visualiztion should be used as often as possible.
>
>Where is DM/KDD being applied or do you expect an opportunity for such
>an application? Marketing? Sales? Finance? Scientific Research? Other?
Most of what I*ve seen has been used to build a foundation for managment
planning. I can imagine data mining being used anywhere, particularlly
in marketing research.
>
>What do consider to be impediments to starting a DM/KDD project?
Data mining is a new evolving specialty area. Perhaps a current problem
is education and awareness, for both the technical and management folks.
>
>Is there a need for a SAS Users Special Interest Group for DM/KDD?
Are you suggesting we data mine the SAS community to see if there is a
group interested in this?
>
>Would you be interested in attending a BOF at SUGI for this purpose?
Sure, thanks for thinking of it and for starting this post.
Tony
==========================================================
Anthony M. Dymond, Ph.D.
Dymond and Associates
A SAS Institute Inc. Quality Partner
4417 Catalpa Ct.
Concord, CA 94521
(510) 798-0129
(510) 680-1312 (FAX)
dymond@ccnet.com
==========================================================
>
>I'd prefer that folks respond to the Discussion Group / SAS-L so that
>others can benefit from your insights.
>
>Thanks!
>
>Tom Skinner
>The views and opinions expressed are those of the author and not
>necessarily those of the Intel Corporation. I am not an Intel
>spokesperson.
>
>
