Which math field is right for me if I'm interested in data mining/machine learning?

[research1]

Hi, I'm sophmore at a state university and I already know that I want to eventually get a PhD and do research/teaching. My question is that if I am interested in eventually doing research in the area of data mining and machine learning, should I focus on pure/applied math courses, statistics or computer science? Which field for grad school would serve me the best for this?

Currently at my university there is a pure mathematician and a computer scientist that have this as a research interest. After talking to both of them about my interest, unsurprisingly the mathematician told me to focus on math and computer scientist told me to focus on computer science. Unfortunately, I was unable to talk to a statistics PhD for their input. Although I haven't researched the field formally, I know that this is an area combining math theory, computing and statistics.

Here's some background on me: I've taken 4 math courses in my freshman year and did quite well. These classes were calculus (integral and multivariate), linear algebra and math probability. I am taking numerical methods, differential equations and a proof class this semester as well. I have only taken one computer science course, but from course descriptions my opinion is that a lot of the undergrad computer science field focuses on software design and architecture. There are also many courses here that I don't think will be of much interest to me: a course on compilers and another on networking for example.

As for pure/applied maths, I think I lean more toward computational applied math. I took a few freshmen seminars that were pure math focused (a primer on abstract algebra and advanced calculus) and I honestly think a lot of the advanced calculus proofs are boring (although I did enjoy linear algebra proofs). Will this be a problem if I pursue a PhD and want to do research in machine learning/data mining? Is teaching with an applied math PhD unheard of? From my observation all the math professors have pure math PhDs and even many of the stats professors have pure math Phds (with a few stats phds here and there).

Finally, I've only taken a math probability course (and the AP stats course) so I have the least experience with the field of statistics. Does anyone know the state of research/teaching in this field?

[CauchyFan]

If you want to do data mining then I would say computer science would be your best bet. A handful of CS courses are just watered down math classes that tell you what you need to know mathwise for CS, so you'll still get the math you need for data mining I would imagine.

Edited September 8, 2010 by CauchyFan

[kowtown]

As the field is interdisciplinary by nature I'd say take a mix of courses from all three, and select a major that allows you to do that. I would lean towards mathematics because not only would you acquire a set of skills but you'd become a better thinker. So perhaps applied or computational mathematics, either of which should allow you to take electives in computer science and statistics. Of course a pure math degree wouldn't hurt either. The rigor of upper-level math courses would likely impress admissions committees more than undergraduate computer science or statistics courses. If someone can grasp real analysis or abstract algebra, they can grasp anything. (Why take a watered-down math course? Though you may find it boring now, an advanced calculus or real analysis course is just good for you.) That should suggest your best course of action: Take a look at machine learning programs at universities you'd be interested in attending and see what they require/request of applicants. Definitely take a look at Carnegie Mellon--they have a department of Machine Learning in their school of computer science. But since you're only a sophomore, you still have a lot of time to figure things out. Best of luck.