From: Nathalie Japkowicz [nat@cs.dal.ca]
Sent: 04 July 2000 17:29
To: Christian Huyck
Subject: Re: Categorisation Issues


Dear Chris,

I think that concentrating on a particular problem such as the data
imbalance problem is a very good idea. It is a problem that many people
encounter and need solutions to.

I am co-organizing a workshop on this problem next month at the AAAI
Conference. The homepage for this workshop is:
http://borg.cs.dal.ca/~nat/Workshop2000/workshop2000.html and you can find
the papers that will be presented at the workshop at: 
http://borg.cs.dal.ca/~nat/Workshop2000/Papers/ 

Other papers that are relevant to this problem are:

Kubat97
@inproceedings{Kubat97,
        Key="Kubat97",
        Author="Miroslav Kubat and Stan Matwin",
        Title="Addressing the Curse of Imbalanced Data Sets: One-Sided
Sampling"
,
        Booktitle="Proceedings of the Fourteenth International Conference
on
Machine Learning",
        Publisher="Morgan Kauffmann",
        pages="179--186",
        Year=1997
}

Kubat98
@article{Kubat98,
 author  = "Miroslav Kubat and Robert Holte and Stan Matwin",
 title   = "Machine Learning for the Detection of Oil Spills in
 Satellite Radar Images",
 year    = 1998,
 journal = "Machine Learning",
 pages   = "195--215", volume=30
}


Ling98
@Inproceedings{Ling98,
   author      ="Charles X. Ling and Chenghui Li",
   booktitle   ="KDD-98",
   key         ="Ling98",
   title       ="Data Mining for Direct Marketing: Problems and
Solutions",
address="",
   year        ="1998"}

Japkowicz95
@inproceedings{Japkowicz95,
author="Nathalie Japkowicz and Catherine Myers and Mark Gluck",
key="Japkowicz95",
title="A Novelty Detection Approach to Classification",
   Booktitle={Proceedings of the Fourteenth Joint
Conference on Artificial Intelligence},
   Year={1995},
        pages =         "518--523",
   Address={},
PUBLISHER={}
   }

A student of mine just finished his Master's thesis on this topic as well,
but I don't yet have the final version of his work. Let me know if you are
interested in the current version of his thesis.

Several of the above papers also refer you to other works on the problem,
so altogether, this should give you a pretty full picture of what has been
done on this problem to this point.

I hope this will be useful, and please, keep in touch! I would be
interested in hearing about further development of your research!

Best Regards,
Nathalie.


--
Nathalie Japkowicz, Ph.D.                    Location after August 8, 2000
Assistant Professor,                         School of Information
Faculty of Computer Science                  Technology & Engineering
DalTech/Dalhousie University                 University of Ottawa      
6050 University Avenue                       http://www.site.uottawa.ca/
Halifax, N.S.
Canada, B3H 1W5

e-mail: nat@cs.dal.ca
WWW: http://borg.cs.dal.ca/~nat
Telephone: (902) 494-3157
FAX:       (902) 492-1517