[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
DM: RE: RE: data mining skeptics
From: Usama Fayyad
Date: Sat, 21 Aug 1999 12:15:22 -0400 (EDT)
this reminds me of a Wired Magazine, late 1994 issue, where they listed "Data Mining" as a hype term and predicted that it will disappear by July 1995... I just came back from KDD-99 which had over 600 attendees, and a great collection of very high quality papers. There is a healthy energetic crowd working on academic, industrial, and applications issues... I think the trick is to remember that for any new area or field, the new solution is typically addressing something people were already doing. People for a long time recognized there is value in their data. What's new here is that the sizes got huge very quickly, so most of the data resource gets wasted... So in a sense, the problem of data mining is not "new". But what is? e.g. when the automobile came along, it was not a revolution in any sense: in fact for a long time it was debatable whether it could ever replace the horse. The same is true for trains, etc. Over time, the extent to which a technology is: - integrated - convenient - easy to use - affordable drives it towards adoption, and after that a 'revolution' might happen. Where data mining seems to be going in the right direction is along these areas. This is in terms of systems, etc. These are directions that, generally speaking, statistics did not try to follow. In contrast, database systems did. So there is a distinction between data mining systems and data mining as a field of study. In the long term, I think data mining will become just another convenient set of tools for reducing data. The more invisible, the more successful (see article by George John in first issue of SIGKDD Explorations http://reseach.microsoft.com/datamine/sigkdd) As a field of study, as far as I am concerned, there are plenty of unsolved problems (see my editorials in the journal Data Mining & Knowledge Discovery for a list) and so many problems that will be around for a long, long time. Data tends to grow, and very fast, and people will be looking for techniques to reduce and exploit it. If data mining as a scientific investigation area goes away, then something pretty much equivalent will replace it. So from a semantics perspective, I am not sure there is any issue here. Researchers will be trying to solve data mining problems for a long time to come. Does it overlap with existing fields? sure. But what field starts out in total isolation anyway. Solid areas of investigaton emerge from other fields. There will always be overlap. The question is where is most of the focus and energy going... Usama. -----Original Message----- From: Tom Dinsmore [mailto:TDinsmore@exapps.com] Sent: Monday, August 16, 1999 6:42 AM To: email@example.com Subject: DM: RE: data mining skeptics There is a great deal of skepticism and cynicism about data mining among businesspeople because the field is overhyped. Marketing executives are subjected to a chorus of sales messages from vendors whose claims are simply ridiculous. Individuals with a great deal of experience in traditional modeling tend to be the most skeptical about new data mining products and services. In some cases, these individuals may perceive that new tools and techniques are a threat to their existing competence and capabilities; if you believe that people are responding defensively to your message, encourage them to think about the advantages and benefits in terms of doing their work better or faster. In other cases, skeptical messages may reflect business values and requirements. For example, data miners tend to focus on very small incremental improvements in model accuracy, improvements that may or may not lead to genuine economic benefits or business utility. Are customers reacting with skepticism to your claims, or are they telling you that they do not value the results you propose to achieve? Finally, bear in mind that it is your customers' job to be skeptical about new products and services -- if data mining survives (and I believe that it will) it is because data miners will succeed in establishing credibility within the domains they seek to serve. -------------------------------------------------------------- Thomas W. Dinsmore firstname.lastname@example.org Engagement Manager Exchange Applications One Lincoln Plaza 89 South Street Boston MA 02111 voice: 617-737-2244 x556 fax: 617-443-9143 mobile: 617-512-7297 -----Original Message----- From: Tjen-Sien Lim [mailto:email@example.com] Sent: Saturday, August 14, 1999 8:07 PM To: firstname.lastname@example.org Subject: DM: data mining skeptics I'm wondering what you'd say to people who are skeptical or even cynical toward data mining. I've just met people who really dislike data mining and predict that the entire field won't last long. Since they're beyond my league, I couldn't convince them that data mining is a real field and won't vanish. Thanks. -- Tjen-Sien Lim (608) 262-8181 (Voice) Dept. of Statistics (209) 882-7914 (Fax) Univ. of Wisconsin-Madison email@example.com 1210 West Dayton Street http://www.stat.wisc.edu/~limt Madison, WI 53706 P.S. My apology if this is a stupid question.