Date: Mon, 29 Mar 1999 15:13:50 CET
From: Berka Petr BERKA@vse.cz
Subject: PKDD99 Discovery Challenge
The Discovery Challenge will be held as a part of the 3rd European
Conference on Principles and Practice of Knowledge Discovery in
Databases (PKDD'99), September 15 - 18, 1999, Prague, Czech Republic.
So only people registered for the PKDD'99 can participate in the
discovery challenge. (For more info about the PKDD conference visit
http://lisp.vse.cz/pkdd99)
The Discovery Challenge should constitute a collection of data and
problems as a common ground for better comparisons and discussions of
the applicability of KDD methods on a real-world problems with respect
to both KDD methodology and application viewpoints.
TASK DESCRIPTION
Once upon a time, there was a bank offering services to private
persons. The services include managing of accounts, offering loans,
etc. The bank wants to improve their services by finding interesting
groups of clients (e.g. to differentiate between good and bad
clients). The bank managers have only vague idea, who is good client
(whom to offer some additional services) and who is bad client (whom
to watch carefully to minimize the bank looses). Fortunately, the
bank stores data about their clients, their accounts (transactions
within several months), the loans already granted, the credit cards
issued etc. So the bank managers hope to find some answers (and
questions as well) by analyzing this data.
The discovery challenge task is to
* define a problem which can help the bank to improve their
services (e.g. define the notion of a good or bad client, suggest
new/current service that can be offered to a group of clients
etc.); from the KDD point of view the problem can be
classification, prediction or description,
* show how KDD can be used to solve the problem (even if the
results are not much significant from the KDD point of view, they
can be important for the bank).
CONTRIBUTIONS
Contributions are expected to present the problem definition, the
necessary preprocessing steps, the applying of data mining algorithms
(either own or third-party approaches) and the interpretation of
results (in terms of both KDD and bank points of view). The main
emphasis of the reported experiments should be on the application
methodology.
Submitted papers should be in English and not exceed 10 single-spaced
pages. The paper must be submitted to Petr Berka by July 1st, 1999. We
prefer electronic submission of PostScript files by e-mail or using
the "submit contribution" option from the PKDD'99 Web homepage. All
contributions will be published in a technical report of the
University of Economics. Wa also think about publishing some
contributions in an international journal.
DISCOVERY CHALLENGE CHAIR
Petr Berka
University of Economics
W. Churchill Sq. 4
CZ 130-67 Prague 3
Czech Republic
Email: berka@vse.cz
phone: +420-2-24095-465
fax: +420-2-793-47-49
Copyright © 1999 KDnuggets