Hello,

I am a masters student and will be starting with my final year master's thesis on Data mining. I am looking for some research areas which are challenging. Also, I am new to DM, so gives me a gradual learning curve as I move along.

Any ideas?

Thanks

asked Sep 02 '10 at 15:05

zengr's gravatar image

zengr
140479

closed Dec 01 '10 at 18:39

The question has been closed for the following reason "Problem is not reproducible or outdated" by zengr Dec 01 '10 at 18:39


2 Answers:

Your ideas should be more personal than something suggested in a web forum. But some ways you might go about getting one:

  1. who's your advisor? You should ask him for areas he works on and related ideas. Doing research that's not in your (good) advisor's comfort zone is a lot harder.
  2. Are there any problems you care about? Any things you always wished should be possible, etc? If so, go read the literature on similar things (and to find out what is the literature on this or related topics you could google scholar a lot of keywords or even ask a question here).
  3. You're interested in data mining, right? Then lookup the archives of KDD, ICDM, and other top conferences in the area. Read a few (10 to 20) papers that catch your eye in the past couple of years. Do you find a method interesting? If so, you can try to apply it to different problems. Do you find a problem interesting? If so, you can look up other methods that can work with it (and here an advisor could really help you).

Are you acquainted with the basic literature in the field? The Witten and Frank book is a classic, and you should read it cover to coverto at least get an idea of what sorts of problems exist and what sorts of techniques can be applied. Books are never state-of-the-art, but they can give you a good starting point.

answered Sep 02 '10 at 19:58

Alexandre%20Passos's gravatar image

Alexandre Passos ♦
2549653277421

my advisor said, read my papers and select a topic and send me a project proposal. Now, I am supposed to read her papers, come up with a topic i "like", write 1 page proposal and mail her. But the problem is, how do I select the topic?! I guess, there is lack of communication b/w me and advisor!

(Sep 03 '10 at 01:18) zengr
1

Maybe there is a lack of communication. Probably what your advisor wants is that you at least have an idea of the things she has worked on, so she can come back at you with how these works can be extended/applied to other areas/etc.

So read her papers and see if there's anything that you would like to learn how to do, or that you find interesting and want to apply somewhere else, etc. Then just tell her that. Don't worry too much about writing a good proposal---writing proposals is an art, and part of it involves consciously not doing all the work right there (a good rule of thumb seems to be that if the final work does exactly what the proposal said it should do then it's because the work itself was uninteresting and trivial, and the author could just have skipped the proposal and written the thesis at once.

(Sep 03 '10 at 06:46) Alexandre Passos ♦

Also maybe you could get some good insights if you asked a question like "how did you come upon your thesis/research subjects" here. Maybe some people have interesting stories to tell.

(Sep 04 '10 at 08:40) Alexandre Passos ♦

I'd start by asking yourself to try and answer these questions (to yourself, primarily):

  • why did you choose to do a masters' degree?
  • why did you choose data mining as your topic?

when you know the answers to these questions, then you should move on and answer these questions (to yourself, but perhaps also over here):

  • did you actually read your advisor's paper? what are they about?
  • did you find any aspect of them interesting? which one? and why?
  • did you find any of your advisor's work to be utterly boring, or utterly bad? if so, why? (this may also be a good starting point)

answered Sep 04 '10 at 10:17

yoavg's gravatar image

yoavg
74192331

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.