Data warehousing LO8684

John Paul Fullerton (jpf@mail.myriad.net)
Sat, 27 Jul 1996 02:43:32 +0000

Replying to LO8675 --

> The remarks about the WWW were particularly interesting,
> and I can see a virtual business adopting a similar model -
> hypermedia is a wonderful medium for exchanging information across
> geographical and temporal spaces.

During the Data warehousing conversation, I visited the Byte magazine WWW
site.

http://www.byte.com

They have a spotlight (same address) with the following introduction.
"Data mining was the subject of one of our most popular State of the Art
sections, in October 1995. Stories included The Data Gold Rush, A Data
Miner's Tools, and Data Mining Dynamite."

The titles link to the articles. The first article mentions data mining as
one way to make use of the data in the warehouse (Oh, watch those
metaphors!) The second and third reference are for quick comments and
related graphics.

Recently at work (at a library), I found a book among the new acquisitions
called "Data Mining with Neural Networks". If anyone will consider an
extended metaphor, I offer the following. On the cover of the book, was a
ladder. Meanwhile, on the announcement display of a church near my house
was the title of a sermon, "Who's climbing Jacob's ladder?" (In the Bible,
angels were climbing the ladder that Jacob saw in a vision. The word angel
means messenger.) So, I reasoned that intelligent agents (computer
programs based on neural network technology) would be climbing around in
the data warehouse finding stuff for people :)

Further explanation of neural networks (not claiming the utmost fluency) -
Use of Neural networks is a method of processing extensive data without
explicit rules for the evaluation of the data. The direction of the
"self-defined" response is developed through "training" the network based
on strengthening responses that fulfill the goals for the system. Possible
uses of such a system in data mining could be "find me an article like
article A". The system accesses data (perhaps identifying grammar, words,
proximity, and such without actually having rules that evaluate the data).
Maybe the system records a "signature" for the file and then seeks other
files with a related "signature". The main idea is that "intelligent"
behavior is produced in a program without the authors of the program
having to define the rules for behavior. It is considered useful where
defining the rules is difficult or impossible or, from another
perspective, where the computer's work in data correlation fulfills
requirements that would have needed more attention from the computer user.

Have a nice day
John Paul Fullerton
jpf@myriad.net

-- 

"John Paul Fullerton" <jpf@mail.myriad.net>

Learning-org -- An Internet Dialog on Learning Organizations For info: <rkarash@karash.com> -or- <http://world.std.com/~lo/>