Sometimes questions are asked, which can be roughly described as:
A: I have an unsorted heap of data (so it is sometimes more, sometimes less required to create predictors)
B: I have a target/response - variable (to create or already available) or a vague idea where I want to get
Resulting question: How do I get from A to B ?
- Logistic Regression and Dataset Structure
- How should I deal with features with time related values in Bayesian network? (which btw should be deleted anytime soon ... the discussion is leading nowhere)
- Failure detection method
- General approaches to model car traffic in a parking garage
- Resources for beginners - how to determine probability of user action based on certain criteria?
I hope you get the point. From my experience on the rapidminer-forum, those questions are invitations for discussions and do not have something like a "correct", but maybe a best answer. As far as I see, answers can only be suggestions to try this or that. Don't get me wrong: I really like this type of questions, because they link practical problems with algorithms/formulas.
- Are those questions generally (or in most cases) cw ?
- How to tag those questions (data-mining seems to be too broad) ?
EDIT: I added two more examples. Additionally, if no one cares to a) either a agree to my proposal or b) deny it and make another one, nothing will be changed. I am just not self-confident enough to accept my proposal and retag all questions of that kind. As far as I understand statsexchange, tags should be discussed instead of created spontaneously.