What main methodology are you using for data mining?
|
Poll |
What main methodology are you using for data mining?
[189 votes total]
|
CRISP-DM (96) |
51% |
SEMMA (22) |
12% |
My organization's (13) |
7% |
My own (43) |
23% |
Other (8) |
4% |
None (7) |
4% |
|
Comments
- Uta Winter, Subject: CRISP-DM
As the ultimate reason of doing Data Mining is always to improve a
business situation which is perceived as being not satisfactory it is
only logical that the Crisp-DM methodology starts there: How do we
tackle the business problem by transforming it into a DM-problem. It
doesn't start with sampling (which is not always necessary anyway)!
This is a big improvement over more techically focused approaches. The
Crisp-Manual is well-written and especially non-tekkies can benefit
from it. After having done some projects myself I can feel that this
process model is not written by people who have never been out in a
company and only discussing DM in theory but who exacly know all the
issues which occur when doing a real project.
- Markus Gretschmann, Subject: Methodology
I did several projects according to CRISP and it always worked fine!
It's the only one that keeps the business goal in mind, not only the
technical aspects! It also helps a lot if you have to communicate
what you do during a project to someone not involved into the process.
- Karl Brazier, Subject: Methodology
Given the choice, I'd be inclined to use my own approach, but in my
experience a problem of trying to apply DM in a research group spawned
from a more traditional area (in my case, electrical engineering) is
that the eminent overseers, lacking a knowledge of DM, try to impose
their traditional approach, which turns out to be quite haphazard from
the viewpoint of systematic DM. I imagine this will be fairly familiar
to others working in groups that are not primarily
DM-focussed. Anyway, I've voted for "other", this being a compromise
between my own system and no system.