(text)
Tom Fawcett, AAAI-98/ICML-98 workshop:
AI Approaches to Time-series Problems http://www.cs.williams.edu/~andrea/aaai98.html
--
Data Mining and Knowledge Discovery community, focusing on the
latest research and applications.
Submissions are most welcome and should be emailed, with a
DESCRIPTIVE subject line (and a URL) to gps.
Please keep CFP and meetings announcements short and provide
a URL for details.
KD Nuggets frequency is 2-3 times a month.
Back issues of KD Nuggets, a catalog of data mining tools
('Siftware'), pointers to Data Mining Companies, Relevant Websites,
Meetings, and more is available at Knowledge Discovery Mine site
at http://www.kdnuggets.com/
********************* Official disclaimer ***************************
All opinions expressed herein are those of the contributors and not
necessarily of their respective employers (or of KD Nuggets)
*********************************************************************
~~~~~~~~~~~~ Quotable Quote ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Prediction is difficult, especially of the future.
Niels Bohr (thanks to http://www.mathacademy.com/ Previous1NextTop
Date: 12 Feb 1998
From: GPS gps
Subject: NY Times on Data Mining Helping Combat Medicare Fraud
As reported by Newsbytes via Individual Inc. :
Medical insurance fraud hit the front pageof The New York Times last week.
NY Times story 'Unwitting Doctors and Patients Exploited in a Vast Billing Fraud'
published Friday, February 6, 1998, Friday in Business/Financial Desk section,
described how government investigators are unraveling a scheme in which phony
medical bills using the names of unsuspecting patients and doctors were
submitted to private insurers.
The scheme was uncovered with the help of data mining technology
developed by IBM, whose Fraud and Abuse Management System
helped uncover the latest racket.
Two observations concerning the 'Barbie Association Rules' problem
(KDNuggets 98:1):
(1) All respondents assumed that customers who buy a Barbie doll have a
higher propensity to buy certain candy bars than customers who do not buy
a Barbie doll -- thus assuming facts not stated. Of course, managerial
action should be based upon a customer segment's relative propensity to purchase, not its absolute propensity to purchase.
(2) The observed association between doll-buying and candy-buying
behavior is itself a product of underlying 'causes' such as consumer
tastes and Walmart's merchandizing mix -- ad spending, promotion and
floor planning, for example. While respondents have different theories
about how best to use this information, all assume that the relationship
will stay the same if the merchandizing mix is changed (which one must
do, by definition, if any action is taken).
The best answer to Mr. Scott's question, IMHO, is that Wal-Mart can test
different treatments, measure the results and draw appropriate
conclusions. The observed association offers useful insight for
hypothesis generation but no self-evident guide to action. This ought to
be the first law of data mining.
Meta Group ranks Silicon Graphics' Mineset third in data mining market
share in its special report titled 'Data warehouse Marketing Trends /
Opportunities' published in January 1998. Mineset is superseded only
by SPSS and SAS, which continue to benefit from their installed
statistical-savvy user bases. MineSet's successful push into this
market in a short time is a demonstration of the program's
unparalleled mining and visualization features.
KDD-98 will provide a venue for a few workshops to focus on advanced
research and development of knowledge discovery, and allow interested
researchers to gather in a relatively small (< 50) group.
Proposals by qualified individuals interested in chairing a workshop
are solicited. Topics include all areas of KDD, including:
the knowledge discovery process, applications in specific areas
(e.g., fraud, churn, financial, bio-chemistry), commercial aspects.
The goal of the workshops is to provide an informal forum for
researchers and practitioners to discuss important issues of current
interest. Concrete open and/or controversial issues are encouraged.
There will be at most THREE workshops sessions in parallel on 31 Aug 1998.
Workshop organizers will have responsibilities including:
1) Writing the call for papers and publishing it.
2) Coordinating workshop participation and content, which involves
arranging short informal presentations and possible panel discussions.
3) Moderating or leading the discussion and reporting its high points,
findings, and conclusions.
4) Writing a brief summary for the AI magazine and/or
Data Mining and Knowledge Discovery (250 words).
Submission Instructions
-----------------------
Interested parties should submit a short proposal for a workshop
of interest by 17 March 1998. The proposal should be e-mailed
to ronnyk@sgi.com
with the subject 'KDD-98 Workshop Proposal.'
Proposals should include a title, organizer(s), a description of the
workshop with objectives, expected number of attendees, tentative list
of invited speakers (if any), and planned format (mini-conference,
panel discussion, or group discussion, combinations of the above,
etc).
Notification of acceptance or rejection will be e-mailed to the
organizer(s) by 24 March 1998. Workshop organizer(s) will then submit
500-word blurbs about their workshops for inclusion in conference
brochure and posting on AAAI web site. The due date for this
blurb is 3 April 1998.
Calls for Papers for accepted workshops will be responsibility of the
organizer(s).
The proposal should motivate why the topic is of interest or
controversial, why it should be discussed and who the targeted group
of participants is. In addition, please send a brief resume of the
prospective workshop chair, a list of publications, and evidence of
scholarship in the field of interest. Submissions should include
contact name, address, e-mail address, phone number, and fax number if
available.
Questions should be addressed to Ronny Kohavi, ronnyk@sgi.com.
A new Exchange Applications white paper entitled 'Increasing Customer
Value by Integrating Data Mining and Campaign Management Software' is
now available from my data mining web page http://www.thearling.com.
The paper's abstract is as follows:
As a database marketer, you understand that some customers
present much greater profit potential than others. But, how will you
find those high-potential customers in a database that contains hundreds
of data items for each of millions of customers? Data Mining software
can help find the 'high-profit' gems buried in mountains of information.
However, merely identifying your best prospects is not enough to improve
customer value. You must somehow fit your Data Mining results into the
execution of marketing campaigns that enhance the profitability of
customer relationships. Unfortunately, Data Mining and Campaign
Management technologies have followed separate paths - until now. Your
organization stands to gain a competitive edge by understanding and
utilizing this new union. This white paper describes how you can profit
from the integration of Data Mining and Campaign Management
technologies.
- Kurt Thearling
Previous6NextTop
From: David Isherwood disherwo@attar.co.uk
Date: Mon, 9 Feb 1998 09:20:33 +0000
Subject: Product announcement - Profiler 4 from Attar
XpertRule Profiler 4.0
Attar Software announces the new release of Profiler data mining
software.
The new High Performance Data Mining Tables (DMT) option enables an
enhanced rule induction mode in Profiler. The result is massive
performance increases when mining using data sources on Windows 95 and
NT platforms. For example, you could build a full decision tree from 1
Million rows/records in times of single minutes. When compared to the
speed of release 3 using memory mode (Profiler - option 2) speed
increases in the order of 100 to 200 times can be expected. The
example was achieved on a standard Pentium 120 with 32MB RAM running
Windows 95.
Also included in version 4 are Association Rule discovery, Clustering
and point and click data transformation.
Also on the Attar site are new data mining case stories from GE
Capital, British Gas, ICI and Carlsberg Tetley. See http://www.attar.com/pages/cases.htm
for a full listing.
Previous7NextTop
Date: Mon, 16 Feb 1998
From: GPS gps
Subject: Knowledge Stream Partners looking for Chief Data Architect
Knowledge Stream Partners, Data Mining Consulting and Integration Company is
looking for a chief data architect.
TASK: Lead the analysis, design and evaluation of data warehouses and
data marts for leading edge data mining and performance support systems.
Lead data warehousing and data mining design team.
The candidates will join a team of world-class experts in data mining and
knowledge discovery and customer management systems.
Requirements: 5+ years of experience with relational database systems
(Oracle preferred).
Experience with very large (>10 Gbytes) data warehouse design in
the database marketing environment is essential.
Excellent understanding of modern methodologies including Data
Modeling, Information Engineering, and Object Oriented
Analysis/Design. Very strong analytical and communication skills.
Some travel, domestic and international, will be required.
We offer very competitive salaries, and our outstanding benefits include
profit sharing, stock options, medical/dental insurance, and a 401(k)
plan.
The Boston branch of the company is conveniently located in
downtown Boston, next to Faneuil Hall, and downtown attractions,
and is accessible by public transportation.
The candidate should be a US citizen or permanent resident or otherwise
authorized for employment in the US.
Please email your resume and a cover letter (in Word format or plain ASCII) to:
Steve Gallant
Senior Scientist
Knowledge Stream Partners
148 State Street
Boston MA 02109
email: sgallant@kstream.com
fax: 617-617-742-5820
SCHOOL OF ELECTRONIC, COMMUNICATION AND ELECTRICAL ENGINEERING
REF: 2577/TECH
RESEARCH ASSISTANT/FELLOW IN VIRTUAL DATA MINING TOOL
Salary stlg 10,018 to stlg 15,411 pa RA/RF
Required in the School of Electronic, Communication and Electrical
Engineering. The primary aim of the work is to investigate the
feasibility of developing a VDMT. You will join a team of
researchers who are the forerunners in establishing the field of
Virtual Data Mining.
The initial phase of the project will be 18 months, starting salary
will depend on experience.
You will have knowledge of virtual reality software tools
(particularly Superscape) and data processing/analysis techniques. A
knowledge of C++ and OO techniques would be an advantage.
CLOSING DATE: Friday 27th February 1998
Application Form and Further Particulars obtainable from the
Personnel Dept, University of Plymouth, Drake Circus,
Plymouth PL4 8AA. Tel: 01752 232168,
E-mail: personnel@plymouth.ac.uk.
Please quote Ref. and Job Title.
Previous10NextTop
Date: Thu, 5 Feb 1998 01:52:54 -0800
From: Ronny Kohavi ronnyk@starry.engr.sgi.com
Subject: Training for Data Mining and Visualization using SGI's MineSet
We are pleased to announce our end-user level course for data mining
and visualization using Silicon Graphics' MineSet product. The course
is geared towards anyone interested in understanding data mining and
visualization with MineSet hands-on experience.
By attending this course, you will understand:
1. Data mining and knowledge discovery.
2. The MineSet product, capabilities, and limitations.
3. How to use MineSet to solve your business problems
and maximize the value of your data.
4. The MineSet interfaces that allow building
applications around MineSet, web-launching, and deployment.
The three-day course is provided by Silicon Graphics' Customer
Education and will be held in
Mountain View, CA starting March 23, 1998.
The classroom is set up with Silicon Graphics workstations to
facilitate hands-on training. The class costs $1125.
Register for the class at http://mineset.sgi.com
under training,
where you can also find more information.
Space for the class is very limited, so register early to ensure
your place. The first class filled in two weeks following our announcement!
Ronny Kohavi, Engineering manager, MineSet.
Maximize the value of your data with data mining and visualization.
Previous11NextTop
From: info@pap.com
Date: Mon, 9 Feb 1998 10:49:11 +0000
Subject: PADD98 Call for Participation
The Second International Conference and Exhibition on
The Practical Application of Knowledge Discovery and Data Mining
2nd Pacific-Asia Conference on Knowledge Discovery and Data Mining
Melbourne, Australia, April 15-17, 1998
C A L L F O R P A R T I C I P A T I O N
Home Page: http://www.sd.monash.edu.au/pakdd-98
PAKDD-98 is pleased to announce that it is now taking registrations.
A copy of the registration form is attached below. The on-line
registration process is fully automated and can be found at http://www.sd.monash.edu.au/pakdd-98/on-line.shtml.
Technical Sessions:
Of the 110 submissions, PAKDD-98 accepted 31 regular papers; an
acceptance rate of 28%. In addition, over 20 papers were accepted as
posters for short presentations. The paper list is available on the
conference home page http://www.sd.monash.edu.au/pakdd-98.
For more information about PAKDD-98 please check the conference's home
page, or contact:
Dr Xindong Wu
School of Computer Science and Software Engineering
Monash University
900 Dandenong Road
Caulfield East, Melbourne 3145
Australia
The purpose of Text Mining can be described as searching what knowledge can
be gathered from a collection of texts, even when the understanding is
imperfect. It does not want to improve on Natural Language Understanding
(it makes use of results obtained in this field, without trying to improve
on them). It tries to increase the amount of knowledge to be extracted for
a given level of understanding (that can be an indexing, etc.).
This workshop aims also at gathering people specialists in
data mining
linguistics
data bases and text (understanding and indexing)
The workshop should succeed in defining precisely what TM is about, and
make clear the differences and commonalities with 'information retrieval'.
Here is a proposal that can serve as first step to the discussions:
1 - Characterize the state-of-the art levels of understanding: indexing,
syntactic analysis, semantic analysis, for instance.
2 - Find the existing tools for each such level, and characterize their
properties.
3 - For each level, define the type of information one starts, in
principle, with.
4 - What knowledge can be gathered from each type of information, and, all
being fixed above, improve on knowledge extraction from each level.
1998 AAAI Fall Symposium on AI and Link Analysis
Orlando, Florida
October 23-25, 1998
Computer-based link analysis is a KDD technique increasingly used
in law enforcement investigations, insurance fraud detection,
telecommunications network analysis, pharmaceuticals research,
epidemiology, and a host of other specialized applications. Link
analysis explores associations among large numbers of objects of
different types. There is both an opportunity and a need to
apply AI technologies to assist human reasoning about complex
networks of relationships. This symposium will bring two
communities into contact: members of the AI research community
who currently have (or could soon develop) useful technologies;
and users of link analysis techniques whose needs go beyond the
capabilities of current software.
Further information about the symposium can be obtained from:
Call for Participation
Joint AAAI-98/ICML-98 Workshop
Predicting the Future: AI Approaches to Time-series Problems
Description
Many dream of being able to predict the future. In finance, accurate
predictions can direct portfolio management decisions. In marketing,
predicting future demand for products and services can direct capital
allocation.
When crystal balls are not available, one may rely on analysis of
historical data to discover predictive patterns. Temporal patterns
are of particular interest because of the large number of high-profile
applications that include historical time series. The goal of this
workshop is to bring together AI researchers who study time-series
problems, along with practitioners and researchers from related fields,
in order to establish common ground.
Topics
We are interested in original research results and application solutions
involving the automated analysis of time-series data.
Authors are asked to address the following questions, where applicable:
* How have you formulated the time-series analysis problem?
Do you build complete classification or regression
models? discover temporal patterns?
* Are you focusing on the creation of a new algorithm? On the
creation of temporally oriented features?
* Have you built on related work from AI or from other communities?
* Is your method designed for a particular application? Do your results
generalize?