Re: DM: Data Quality Metrics

• Organization: LIACC/FEP

Dear Rebecca,

 > This would give us an
 > ranking of the overall "quality" of the data, which in turn could suggest
 > where we should put most of our effort into data cleaning or whether the
 > data set is suitable for data mining at all.

A complementary approach is being developed in the METAL project
(http://www.cs.bris.ac.uk/~cgc/METAL/). Data characteristics together with
information about the performance of data mining algorithms are used to give
advice on which algorithm/pre-processing method should be used on a given
problem.

Some of the work developed in this project will be presented at ICML as well as
one of the associated workshops ("What Works Well Where?" -
http://www.research.ibm.com/MetaLearning/).

If you would like to know more about this, don't hesitate to ask.

 > Thank you very much for your time.  If the responses warrant, I will
 > summarize them and post them to the list for everyone to see.

Yes, that would be very interesting.

Cheers,
Carlos

--
Carlos Soares
FEP/LIACC, University of Porto, R. Campo Alegre, 823, 4150-180 Porto, Portugal
EMAIL: csoares@ncc.up.pt - WWW: http://www.ncc.up.pt/~csoares
PHONE: +351 226078830 - FAX: +351 226003654

• Prev by Date: DM: RE: Classification AND Association
• Next by Date: DM: SAS Institute Hosts M2000 - 3rd Annual Data Mining Technology
• Prev by thread: DM: Re: Data Quality Metrics
• Next by thread: DM: cfp: PKAW2000 in Sydney, Australia
• Index(es):
  • Date
  • Thread