KDD Nuggets Vol 1, No. 8 -- 10 December, 1993 
		--- Happy Holidays ---

	KDD-94 workshop will take place at AAAI -- full CFP will
	be distributed soon.
Contents: 
	* Gregory Piatetsky-Shapiro: brief reviews of articles on 
	Intellisense, 
	Fraud Detection Systems, 
	Multi-Dimensional Data Models, 
	"Sucker" Databases, and
	Extracting Knowledge from Diagnostic Databases

-- This issue contains brief reviews of articles I found interesting.
Similar reviews are most welcome. Volunteers ?

-------------
The KDD Nuggets is an informal list for the dissemination of
information relevant to Knowledge Discovery in Databases (KDD), such
as announcements, reviews, and descriptions of conferences, 
workshops, tools, and applications. Strong opinions and informal 
anecdotes are also welcome. 

Contributions to kdd@gte.com; Add/delete requests to kdd-request@gte.com 
 -- Gregory Piatetsky-Shapiro 

--------------------------------------------------
From: Gregory Piatetsky-Shapiro (gps@gte.com)
Subject: Some interesting articles 

MICROSOFT INTELLISENSE:
Oct 1993 Intelligent Systems Report (D. Blanchard Editor, Atlanta, GA)
reports that Microsoft has announced a new feature, Intellisense, that 
will add intelligent capabilities to such software products as Microsoft
Word, Excel, Project, and Powerpoint.  Intellisense embodies a heuristic
programming language that allows it to monitor the user's work patterns
to suggest more efficient ways to use the program.
 
NEURAL-BASED FRAUD DETECTION SYSTEM:
 The same issue also reports that Nestor (Providence, R.I.) has signed
agreements with several banks to install its neural-based credit card 
Fraud Detection System. The FDS software learns a cardholder pattern
of use and is capable to detect changes in these patterns that might
indicate fraud.  The system catches changes as they begin to occur,
providing the bank with the earliest warning of fraud.  FDS systems are
installed at Mellon Bank (Pittsburgh) and in Sligos SA, a French-based
financial services firm.

MULTI-DIMENSIONAL DATA MODEL: (thanks to Chris Matheus for the article)
October 1993 IEEE Spectrum has an article by J. Stamen, on
"Structuring databases for analysis" (pp. 55-58). He presents a
multi-dimensional model (MDM), where the basic unit is a dimension
(e.g. time, product, region).  The data is structured by these
dimensions into a multi-dimensional array.  Each dimension exists only
once for the entire database, ensuring referential integrity.  Stamen
argues that MDM systems are more suitable than relational ones for
many kinds of data analysis, such as computing lag or change in a
field -- operations which would require complex joins in a relational
model.  Stamen is the president of IRI Software, a developer of MDM
systems.

DATABASES of SUCKERS: 
Nov 22, 1993 ComputerWorld article, "Wanna buy a sucker?", by A.
LaPlante, pp. 77-81, describes how some companies create and exploit
databases of so-called "suckers" -- people who are considered likely to buy 
from direct mail.  Such lists include people who responded to 
various sweepstakes and get rich schemes, people who were 
turned down for credit, people who called 900 numbers to claim
instant prizes, etc.  The article describes several cases of prosecution 
for mail fraud and suggests guidelines to companies which want to sell
their lists ethically. 

EXTRACTING KNOWLEDGE from DIAGNOSTIC DATABASES: 
This article in Dec 1993 IEEE Expert, by R. Uthurusamy et al (GM
Labs), describes a very interesting system that uses machine learning
techniques to acquire diagnostic knowledge automatically from the
trouble reporting databases.  Initially, the system uses natural
language analysis to convert user input into a standard form, to which
decision-tree method Inferule was applied. Comparisons show that
Inferule gives better results than ID3 on inconclusive data (i.e. data
which is insufficient for perfect classification).  The system is now
being reviewed for possible deployment.
--------------------------------------------------