The Challenge is
Growing
According to some recent studies that have coveredon the tremendous growthincrease inin the amount of data being recorded and stored on
digital media, it is clear that the gap between this digitally stored datathe amount of data going to
storage, and our ability to
process itit is widening. The
primary driver for this phenomenon is the
a more aggressive relative of the familiar “Moore’s
Law”. Moore’s Law is the observation
that the computational power seems to doubles approximately every
18 months. The corresponding law in the world of digital storage is even more
amazing. Storage capacity, for a fixed price, appears to be doubling
approximately every 9 months! [1]. In a
world governed by these two empirical laws, the importance of the it becomes very clear why the fields of databases and of
Data Mining and Knowledge Discovery are growing in importance. From
a strategic perspective, our need to be able to navigate
and exploit the rapidly growing
universe of digital data will rely heavily on our ability to come up with the
right systems that can
scale towith thesethese growth rates, of growth and that allowing us to effectively
manage and mine the raw data.
The Data Mining field
is centers about developing algorithms for extracting
structure from data. Theis structure can take
the form of statistical patterns, models, and relationships
over the data. Viewed at a high level, these “structures”
are basic data reductions that enable us to summarize the data, to bring
it down in dimensionality to spaces thatin which we have some hope of
understanding and visualizing. These
structures provide the basis for, and to enable us to predict and
anticipate when certain events occur. When viewed at this level, one begins to
understand the fundamental importance of data mining in the future of the
digital data universe. For a more futuristic outlook, see the article [2].
![Text Box:
DILBERT reprinted by permission of United Feature Syndicate, Inc. [2000].](./image002.gif)
Beyond
the Giants and the Monsters
A recent study
conducted at UC-Berkeley indicates
that humankind seems to be producing over two exabytes of information per year. [3]. This study confirms
the persistence of the original trends first uncovered in [4]. The amount of
information generated in 1999 constitutes more than 12% of the aggregate amount
of information generated by humankind in all of recorded history. We conjecture that the rate is more dramatic for the year 2000. I have not seen the computation done for the year 2000, but I am
sure that the rate is getting even more dramatic. It is interesting that thean overwhelming majority
of the collected data is never seen
by humans [4].
As recently as three years ago, people were
wondering how they willwould manage terabyte
databases. Jim Gray, who is familiar with the
systems and scalability issues ofwhen dealing with dealing
with very large databasesamounts of data, used to refer to
these databases as “Terror Byte”
databases. It is interesting that now we
have now gone beyond the gigabytes and the terabytes, and are now
worried about petabytes and exabytes. Interestingly, “Giga” derives from the
Latin term for “Giant”, and “Tera” derives from the
Greek term for “Monster”. A gigabyte is 109 bytes, while a terabyte is
1012
bytes. The next higher units up become have less colorful in names: Ppetabyte and Eexabyte. Interestingly, tThe next term is units up the the Zzettabyte (deriving
from the last letter of the Greek alphabet), and
then the Yyottabyte, indicating a
backtracking from going to the last letter.
In the Data Mining
field, it is fairly safe to think that we are wrestling with the issues ofin dealing with giants (gigabytes). We certainly have
note stepped up to the Mmonsters yet. It is worrisome that within this plethora of
names , we still worry that we barely have
anyfew ideas on how to to even approach the monsters. The terabytes are still indeed the “terror bytes”. However, given the
driving forces of technology out there, it is with urgency that we must start to see progress
in scaling to the larger
databases. With each year, the challenges will grow larger, and so will the corresponding pressure for
breakthrough advancement in data mining technologies,. Data
mining solution providers will be measured along the lines of successful real world applications. Integrating
with new generation
business practices is essential.in modes of providing data
mining solutions, and in real examples of their integration with the way the
new generation of business is conducted.
A large number of cCorporations currently are and have been collecting their business data. and the volume of this data
grows significantly every year.
Furthermore, with the spreadgrowth of E-commerce and the
World Wide Web, online stores are automatically collecting huge amounts of valuable data automatically. For
example, web servers typically generate large volumes of daily access logs data, that capturescapturing page the sequences
of page accesses for millions of
users. Sites can easily generates tens of gigabytes of
data per day. Thus, we can
easily witness the wide availability of huge amounts of data; data that was hardly
there three or four years ago. And the volume of this data grows significantly
every year.
However, it is very difficult to turn such data
into useful information due to the limited availability of scalable data mining
algorithms. For the algorithms
to be scalable, the execution time should increase linearly, or preferably
sub-linearly,
in proportion to the size of the input data. Traditional data analysis
algorithms generally do not scale up well
with limited system resources such as (e.g. main memory). Therefore, we are
now faced with exciting opportunities and great challenges for data mining to
develop efficient, scalable and efficient algorithms that are acceptable in large databases.
This “scalability challenge” goes beyond just the algorithms to include: . It includes how systems issues
are addressed, the delivery of serviceshow services and capabilities
are delivered, and the delivery of mining
results. Results how the results of mining areneed to be made usable and
useful for individuals in businesses
and organizations who knowing little about data mining, statistics, and other esoteric
technologies. This
challenge must be addressed for data mining to cross into its next level of
integration within mainstream business and the
daily operations of organizations.
Special Issue on “Scalable Data Mining Algorithms”
We focus this issue of
SIGKDD Explorations on the topic of Scalable Data Mining Algorithms, and we
welcome Kyuseok Shim as Guest Editor. On this challenging and exciting topic, eleven articles are
included in the section
on contributed
article section on scalable data
mining algorithms.
Kristin P. Bennett and
Colin Campbell contributed a nice tutorial article on support vector machines
that have become popular tools for data mining applications. Jiawei Han and
Jian Pei present a new methodology called frequent pattern growth that avoids
candidate generation in Apriori-style algorithms. This technique can be an
alternative solution for mining frequent patterns when the candidate generation
and test cost is very expensive in applications with long and numerous
patterns. Laks. V.S. Lakshmanan, Carson Kai-Sang Leung, and Raymond Ng describe
a data structure called segment support map that boosts the performance of
frequent itemset mining algorithms by computing sharper upper bounds on the
support and better exploitation of constraints. Wei Wang, Jiong Yang and Philip
S. Yu propose several models to address different types of noise and provide
scalable algorithms for each model. George Forman and Bin Zhang present a
technique for developing parallel versions of centroid-based clustering algorithms that
communicate only
through
sufficient statistics. The aApproaches for
exploiting and pushing user-defined model constraints to speed up the mining
process are discussed by Minos Garofalakis and Rajeev Rastogi. Such techniques
can help to improve the system performance significantly by focusing on only
interesting patterns by users. Thus, it is worthy of investigation for many
existing data mining algorithms. Claudia Diamantini and Maurizio Panti show
that compression capabilities of the Vector Quantizer architecture can be
exploited to find simple and effective classification rules when it is trained withusing a stochastic gradient
algorithm. William A. Maniatty and Mohammed J. Zaki address desirable design
features of parallel data mining algorithms and investigate a migration path
from sequential algorithms by considering an integrated hardware and software
approach. Yves Bastide et al. present the PASCAL algorithm, thatwhich improves Apriori -style algorithms based on pattern counting
inference thatand can reduce database
access significantly. Gregory Piatetsky-Shapiro and Sam Steingold propose a new
measure called L-quality that helps to compare predictive models by looking at
the entire lift curve. Stephen D. Bay et al. describe new data sets that are available
in the UCI KDD archive to help in evaluating scalable data mining algorithms.
Considering it is very difficult to get large real-life data sets to conduct
performance studies of scalable algorithms, this addition of large data sets to
UCI KDD Archive will stimulate and foster systematic progress significantly in
developing scalable data mining algorithms.
KDD-2000 Reports
Also included in this
issue of Explorations are a series of reports from KDD-2000: The Sixth ACM SIGKDD International
Conference on Knowledge Discovery and Data Mining. KDD-2000 took place in Boston, MA on August 20 – 23, 2000. The organizing committee expected 700
attendees. After the final numbers were
tabulated, a total of 920 people attended
KDD-2000! Enthusiasm
and interest in the field is rapidly increasing.
The KDD Cup 2000
annual data mining competition consisted of five questions to be answered based
on web-browsing data collected from an online retailer. Ron Kohavi, Carla Brodley, Brian Frasca,
Llew Mason, and Zijan Zheng provide an excellent report on the competition including
summary reports by the winners of the five question areas.
KDD-2000 conference
highlights included workshops and panel presentations. I thank the
organizers for the time and effort required to provide reports from their
respective workshops and panels. Marko Grobelnik, Dunja Mladenic, and Natasa Milic-Frayling have
provided a report on the KDD-2000 Workshop on Text Mining. Simeon Simoff and Osmar Zaïane report on
MDM/KDD2000: The 1st
International Workshop on Multimedia Data Mining. A summary of the WEBKDD 2000
workshop was provided by Myra Spiliopoulou, Jaideep Srivastava, Ron Kohavi and
Brij Masand. A report on the Workshop
on Distributed and Parallel Knowledge Discovery by Hillol Kargupta, Joydeep
Ghosh, Vipin Kumar, and Zoran Obradovic is included. Ivan Bruha and A. Famili have provided a report on the Workshop
on Postprocessing in Machine Learning and Data Mining. Alexander Tuzhilin summarized the Panel on
Personalization and Data Mining.
We thank the organizers for the
time and effort required to provide these reports.
Concluding Remarks
With this issue of SIGKDD Explorations we launch our second volume of the newsletter. With our continued growth and
the continued growth of the field we need to grow Explorations accordingly. We are pleased to bring in two
changes to the editorial staff and format of your newsletter. We have invited
Paul Bradley to join the editoraleditorial staff as Associate Editor. We have also asked him to serve as Guest Editor
of this issue. This represented the second change in editorial format: future
issues will have themes and special topics with guest editors invited to cover each special topic. We shall of
course continue to include contributed articles on topics of interest and
conference and workshop reports. In
addition to serving as Guest Editor on this issue, Paul has also taken a major
load of additional work on producing this issue in his new role as Associate
Editor. Having welcomed Paul
aboard, we’ll let him introduce this issue’s special focus.
The
special issue section of
this issue SIGKDD
Explorations is focused on “Internet
Data Mining”. A large number of
corporations have been collecting their business data and the volume of this
data grows significantly every year. Furthermore, with
the spread of E-commerce and
World Wide Web, online stores are collecting huge amounts of
valuable data automatically. For example, web servers typically generate large volumes of daily access log data that captures the sequences of page accesses for
millions of users. Thus,
we can easily witness the wide availability of huge amounts of data that was
very difficult in the past. However, it is very hard to turn
such data into useful information due
to the limited availability of scalable data mining algorithms. For the algorithms to be scalable, the execution time
should be increased almost linearly in proportion to the size of the input
data. Traditional data analysis algorithms generally do not scale up well with
limited system resources such as main memory. Therefore, we are now faced with exciting
opportunities and great challenges for data mining to develop scalable and efficient algorithms
that are acceptable in large databases. We focus this issue of SIGKDD Explorations on the topic of “Scalable Data Mining Algorithms”.
The “Internet”
continues to provide the general population of computer users with innumerable
benefits (and possibly some drawbacks).
It also provides the data mining community with large amounts of data
and great challenges through new applications areas like customer relationship
management, e-commerce and web search.
As
the internetInternet expands, there is a
greater need to effectively index and classify content to support efficient
search and navigation. In addition, the
internetInternet provides businesses
with an operational platform for interaction with their customers 24
hours per day, 365 days per yeararound the clock,
without geographic or physical boundaries.
According to a Forrester report, “worldwide net commerce – both B2B and
B2C – will hit $6.8 trillion in 2004” [3].
Databases currently are (and will continue to be) critical components in
the e-commerce arena. The data
generated by this domain provide “all the right ingredients for successful data
mining” [1]. In addition to providing
some of the key foundations in internet content indexing, text mining
techniques may prove very useful in helping the end-user deal with the
information overload streaming down to her or him by processing email and other
electronic communication and prioritizing messages and tasks
[2].
While there is such
excitement enthusiasm and
opportunity for data mining in the internetInternet
arena is
exciting and intriguing,
the serious issue of privacy needs to be effectively handled. Although it is not only an issue with data
mining and the internetInternet, data mining
practitioners in this space should need to be
constantly aware of such the implications of tracking and
analysis technologies on privacy. Without properly addressing the issue of
privacy on the internetInternet, this abundant tap of
data may eventually flow much slower due to regulations, etcand other corrective
or preventive restrictions.
GGiven
the opportunities, challenges and issues touched upon above, we chose to focus this
issue of SIGKDD Explorations focuses
on the topic of “Internet Data Mining”. On this challenging and
exciting topic, five eleven articles
are includedcontributed article section on scalable data miningspecial algorithms focus section. Kristin P. Bennett and Colin Campbell
contributed a nice tutorial article on support vector machines
that have become popular tools for data mining applications. Jiawei
Han and Jian Pei present a
new methodology called frequent pattern growth that avoids candidate generation in Apriori-style
algorithms. This technique can be
an alternative solution for mining frequent patterns when the candidate
generation and test cost is very expensive in applications such as long and
numerous patterns. Laks.
V.S. Lakshmanan, Carson Kai-Sang Leung, and Raymond Ng describe a data
structure called segment support map that boosts the performance of frequent itemset mining algorithms by computing sharper upper bounds on the support and better exploitation of
constraints. Wei Wang, Jiong Yang
and Philip S. Yu propose several
models to address different types
of noises and provide scalable algorithms for each model. George
Forman and Bin Zhang present a technique for developing parallel version of centroid-based clustering algorithms that
communicate only sufficient statistics. The scalable algorithms for exploiting and pushing user-defined model constraints to speed up during mining process are discussed by Minos Garofalakis and Rajeev
Rastogi. Such techniques can help to improve the system
performance significantly by focusing on only interesting patterns by users. Claudia Diamantini and Maurizio Panti show that
compression capabilities of Vector Quantizer architecture can be exploited to
find simple and effective classification rule when we train it with a stochastic gradient algorithm. William
A. Maniatty and Mohammed J. Zaki address desirable design features of parallel data mining algorithms and investigate migration path from sequential
algorithms by considering an
integrated hardware and software approach. Yves
Bastide et al. present PASCAL algorithm that improves Apriori style algorithms
based on pattern counting inference that can reduce database access
significantly. Gregory Piatetsky-Shapiro and Sam Steingold propose a new measure called L-quality that helps
to compare predictive models by
looking at the entire lift curve.
Stephen D. Bay et al. describe
new data sets that are available in the UCI KDD archive to help
the test of scalable data mining algorithms. Considering it is very difficult
to get large real-life data sets to conduct performance study of scalable
algorithms, this addition of large data sets to UCI KDD Archive will stimulate
and foster systematic progress
significantly in developing
scalable data mining algorithms.
Also included in this
issue of Explorations are a
series of reports from
KDD-2000: The Sixth ACM SIGKDD International Conference on
Knowledge Discovery and Data Mining.
KDD-2000 took place in Boston, MA on August 20 – 23, 2000. The organizing committee expected 700 attendees. After the
final numbers were tabulated, a total of 920 people attended KDD-2000! Enthusiasm and interest in the field is rapidly increasing.
The KDD Cup 2000 annual data mining competition consisted of five
questions to be answered based on web-browsing data collected from an online
retailer. The
report on the KDD 2000 Cup by Ron
Kohavi, Carla Brodley, Brian
Frasca, Llew Mason, and Zijan Zheng includes summary reports by the winners of the five
questions.
Conference highlights
included workshops and panel presentations.
We thank the
organizers for the time and effort required to provide reports from their
respective workshops. Marko Grobelnik, Dunja Mladenic, and Natasa Milic-Frayling have provided a report on the KDD-2000 Workshop on
Text Mining. Simeon Simoff and Osmar Zaïane report on MDM/KDD2000: The 1st
International Workshop
on Multimedia Data Mining. A summary of the WEBKDD 2000 workshop was provided
by Myra Spiliopoulou,
Jaideep Srivastava, Ron Kohavi and Brij Masand. A report on the Workshop on Distributed and
Parallel Knowledge Discovery by Hillol Kargupta, Joydeep Ghosh, Vipin Kumar, and Zoran Obradovic is included. Ivan Bruha and A. Famili have provided a report on the Workshop on
Postprocessing in Machine Learning and Data Mining. Alexander Tuzhilin
summarized the Panel on Personalization and Data Mining.
Paul Bradley, please fill this part!!
Kosala
and Hendrik Blockeel contribute a survey on web mining research, providing an
overview of sub-areas and problems encompassed by web mining. Yiming Ma, Bing Liu and Ching Kian Wong
propose a web-browser-based graphical user interface for displaying, navigating
and annotating a set of discovered rules.
Sunita Sarawagi and Sree Hari Nagaralu discuss the notion, use and
implication of providing data mining models as services on the web. Stanley Loh, Leandro Krug Wives and José
Palazzo de Oliveira propose a concept-based approach to analyzing texts
extracted from the web using an interesting collaboration between human and
computer to obtain the final result.
José Borges and Mark Levene present a heuristic for speeding up a
previous system based upon a hypertext probabilistic grammar for extracting
browsing behavior from web log data.
Additional articles addressing general topics of
interest to the SIGKDD community include the introduction of a simple new
scalable k-means clustering algorithm proposed and evaluated by Frederik
Farnstrom, James Lewis and Charles Elkan.
We hope that this article prompts further discussion on scalable data
mining algorithms in general. A survey
and overview of algorithms discovering association rules appears by Jochen
Hipp, Ulrich Güntzer and Gholamreza Nakhaeizadeh, including several runtime
experiments. Alex Freitas provides a
position paper outlining differences between the classification task and that
of discovering association rules and raises many interesting points, hopefully
prompting further discussion on the subject from the SIGKDD community.
.
Reports from workshops and meetings of interest
include a report from Jeanne Behnke, Elaine Dobinson and others on the NASA
Workshop on Issues in the Application of Data Mining to Scientific Data. Coming from one of the early organizations
with a definite need to analyze large data sets, the topics discussed in this
report provide interesting insight into the relationship between the data
analyst and the domain specialist in scientific areas of interest to NASA. Dan Suciu and Gottfried Vossen have kindly
provided a report from WebDB’2000 following the ACM PODS/SIGMOD
conferences. Dimitrios Gunopulos and
Rajeev Rastogi have, similarly, kindly provided a report on the ACM SIGMOD
Workshop on Research Issues in Data Mining and Knowledge Discovery.
We hope that SIGKDD
Explorations continues to meet your expectations on covering timely topics and
events in this growing field. Our shift in format towards special focus topics
is not intended to limit the scope of our coverage of the field. We still welcome
articles and contributions, including position and idea papers as well as
controversial topics that stimulate discussion and deliberation. We hope the
special focus sections will allow us to continue the general breadth of
coverage while providing some additional depth along the focus topic.
We usrge interested readers and
contributors to propose special topics. We welcome guest
editors from within the field and from outside it to help us cover related
focus topics. Please contacts
us with any iideas for future topics
of interest, and any feedback on Explorations.
We conclude with a reminder that the premier event
in the SIGKDD community is quickly approaching. KDD-20010 is
taking place in BostonSan Fransisco, MACA,
beginning Sunday, August 260th
and concluding on August 23rd. See the
conference website for details (http://www.acm.org/sigkdd/kdd20010). We look forward to seeing you in Boston!
KDD-2001 in San
Francisco, California, USA
IWe conclude with a
reminder that the premier event in the SIGKDD community, KDD-2001, will take
place in San Franscisco, CA,
beginning Sunday, August 26th and concluding on August 29th
next year. TheConference details of the conference including paper submissions
are available at http://www.acm.org/sigkdd/kdd2001. We are looking forward to seeing you in San Franscisco!
Acknowledgement: tThe guest editor would
like to thank Paul Bradley, Usama Fayyad and Sunita Sarawagi for the
opportunity of contributing to SIGKDD Explorations. Without the support of
Yesook Shim, it would have been impossible to complete the special focus sections
of this issue.
References
[1] J. Gray and P.
Shenoy. Rules of thumb in data engineering. in Proceedings of 16th
International Conference on Data Engineering, pages 3-12. IEEE, 2000.
[2] U. Fayyad. “The
digital physics of Data Mining”, Communications of the ACM, March 2001.
[3] P. Lyman, H. R.
Varian, J. Dunn, A. Strygin, K. Swearingen, 2001. “How Much Information”,
University of California, Berkeley, Technical Report.
[4]
M. Lesk. “How much information is there in the world?” Technical report,
http://www.lesk.com/mlesk/ksg97/ksg.html 1997.
About the Editors
About the Editors
REFERENCES
[1]Kohavi,
R. Mining E-commerce Data: Challenges and Stories from the
Trenches. Presentation at DIMACS/IBM
Workshop on Data Mining in the Internet Age.
May 2, 2000.
[1]Markoff,
J. Microsoft Sees Software ‘Agent’ as
Way to Avoid Distractions. New York
Times, Technology Section, July 17, 2000.
[1]Sanders,
M. R., and Temkin, B. D. Global
eCommerce Approaches Hypergrowth. The
Forrester Brief, April 18, 2000.
http://www.forrester.com.
Editor-in-Chief:Paul Bradley is a Researcher in
the Data Management, Exploration and Mining Group at Microsoft Research. Research interests include classification
and clustering algorithms; underlying mathematical problem formulations; and
issues related to scalability. After
receiving the Ph.D. degree from the University of Wisconsin-Madison in 1998 on the topic of mathematical
programming and data mining, he joined Microsoft Research where he works on research in developing new data mining algorithms as well as on shipping data mining components in
Microsoft products such as SQL Server and Commerce Server.
Usama Fayyad is President & CEO
of digiMine, Inc..
He
received his Ph.D. from The University of Michigan, Ann Arbor in 1991. He is an
Editor-in-Chief of Data Mining and Knowledge Discovery, the primary
technical journal in the field, and has chaired several past KDD conferences.
Prior to digiMine, he was at Microsoft where he founded and led Microsoft
Research’s Data Mining & Exploration (DMX) Group. His work with Microsoft product groups included the development
of data mining prediction components that ship with Microsoft Site Server
(Commerce Server 3.0 and 4.0) and developing scalable algorithms for mining
large databases and architecting their fit with server products such as
Microsoft SQL Server and OLAP Services.
In addition to managing the DMX research group, he managed the core part
of the development team that is building providers in the SQL Server product
group. He was also a driving force
behind establishing a new industry standard in data mining based on Microsoft’s
OLE DB API. Prior to joining Microsoft, Usama was at the Jet Propulsion
Laboratory (JPL), California Institute of Technology (1989-1995) where he
founded and headed the Machine Learning Systems Group and developed data mining
systems for the analysis of large scientific databases. For this work he
received the most distinguished excellence awards from Caltech/JPL and a U.S. Government
Medal from NASA. He remained affiliated with JPL as Distinguished Visiting
Scientist after he moved to Microsoft.
Associate Editors:
Sunita
Sarawagi is assistant professor at the Indian Institute of
Technology, Bombay. She received her
Ph.D. degree in 1996 from the University of California at Berkeley and was a
research staff member at the data mining group of IBM Almaden Research Center
before joining IIT Bombay in 1999. Her
research interests include database mining, OLAP and extended relational
database systems. She has several
publications in international conferences including a best paper award at the
1998 ACM SIGMOD conference. She has
served as program committee member for SIGMOD, VLDB and ICDE conferences and is
an associate editor of the ACM SIGKDD newsletter and IEEE Data Engineering
Bulletin.
Paul
Bradley is a developerData Mining Engineer at digiMine, Inc. His focus is on the integration and architecture of
data mining services. He received his
Ph.D. from the University of Wisconsin in 1998 on the topic of mathematical
programming and data mining. Prior to
joining digiMine, he was a researcher in the Data Management, Exploration and
Mining Group at Microsoft Research.
Research interests include classification and clustering algorithms;
underlying mathematical programming formulations; and issues related to
scalability. While at Microsoft
Research, he worked on data mining components shipping with SQL Server and Commerce Server.
Sunita Sarawagi is assistant professor
at the Indian Institute of Technology, Bombay.
She received her Ph.D. degree in 1996 from the University of California
at Berkeley and was a research staff member atwith the data mining group
of IBM Almaden Research Center before joining IIT Bombay in 1999. Her research interests include database
mining, OLAP and extended relational database systems. She has several publications in
international conferences including a best paper award at the 1998 ACM SIGMOD
conference. She has served as program
committee member for SIGMOD, VLDB and ICDE conferences and is an associate
editor of the ACM SIGKDD newsletter and IEEE Data Engineering Bulletin.
Guest Editor:
Kyuseok Shim is currently an
Assistant Professor at Korea Advanced Institute of Science and Technology
(KAIST) in Korea. Before joining KAIST, he was a member of technical staff
(MTS) in Bell Laboratories in which he initiated and worked for the Serendip
data mining project with Rajeev Rastogi. Before that, he was a research staff
member of Quest Data Mining project at IBM Almaden Research Center. He also
worked as a summer intern for two summers at Hewlett Packard Laboratories. He
received B.S. degree in Electrical Engineering from Seoul National University
in 1986, and the MS and Ph.D. degrees in Computer Science from University of
Maryland, College Park, in 1988 and 1993 respectively. He has been working in the area of data mining, data warehousing, query
processing and query optimization, XML and semi-structured data. He has been an
Advisory Committee Member for ACM SIGKDD. He has also served as a program
committee member on international conferences that include ICDE'97, KDD'98,
SIGMOD'99, SIGKDD'99, VLDB'00 and SIGKDD’01. He did a data mining tutorial at
CIKM'99, ICDE'99 and SIGKDD'99 conferences.
He was co-chair of the 1999 ACM SIGMOD Workshop on Research Issues in
Data Mining and Knowledge Discovery.
Kyuseok Shim is currently an Assistant
Professor at Korea Advanced Institute of Science and Technology (KAIST) in
Korea. Before joining KAIST, he was a member of technical staff (MTS) in Bell
Laboratories in which he initiated and worked for the Serendip
data mining project with Rajeev
Rastogi. Before that, he was a research staff member of Quest Data Mining
project at IBM Almaden Research Center. He also worked as a summer intern for
two summers at Hewlett Packard Laboratories. He received B.S. degree in
Electrical Engineering from Seoul National University in 1986, and the MS and
Ph.D. degrees in Computer Science from University of Maryland, College Park, in
1988 and 1993 respectively.
He has been working in the area of data mining, data
warehousing, query processing and query optimization, XML and semi-structured
data. He has been an Advisory Committee Member for ACM SIGKDD. He has also
served as a program committee member on international conferences that include ICDE'97, KDD'98,
SIGMOD'99, SIGKDD'99, VLDB'00 and
SIGKDD’01. He did a data mining tutorial
at CIKM'99, ICDE'99 and SIGKDD'99 conferences.
He was co-chair of the 1999 ACM SIGMOD Workshop on Research Issues in
Data Mining and Knowledge Discovery.
Usama Fayyad is President & CEO
of digiMine, Inc..com. He received his Ph.D. from The University of
Michigan, Ann Arbor in 1991. He
is an Editor-in-Chief of Data Mining and Knowledge Discovery, the primary technical journal in the field, and has
chaired several past KDD conferences. Prior
to digiMine, he was at Microsoft where he founded and led Microsoft Research’s
Data Mining & Exploration (DMX) Group.
His work with Microsoft product groups included the development of data
mining prediction components that ship with Microsoft Site Server (Commerce
Server 3.0 and 4.0) and developing scalable algorithms for mining large
databases and architecting their fit with server products such as Microsoft SQL
Server and OLAP Services. In addition
to managing the DMX research group, he managed the core part of the development
team that is building providers in the SQL Server product group. He was also a driving force behind
establishing a new industry standard in data mining based on Microsoft’s OLE DB
API. Prior to joining Microsoft, Usama was at the Jet Propulsion Laboratory
(JPL), California Institute of Technology (1989-1995) where he founded and
headed the Machine Learning Systems Group and developed data mining systems for
the analysis of large scientific databases. For this work he received the most
distinguished excellence awards from Caltech/JPL and a U.S. Government Medal
from NASA. He remained affiliated with JPL as Distinguished Visiting Scientist
after he moved to Microsoft.