Is it okay to work on a problem which is incremental as your first Ph.D problem?

Yes!

Your goal for your first problem is just to publish something. It doesn’t have to be significant. It doesn’t have to be in a top conference. It doesn’t have to be related to your eventual thesis topic. It doesn’t even have to be in the same subfield as anything else you publish in the future. Just publish something.

This is just getting you used to the process of writing and revising and submitting and being frustrated by bad reviews and rerevising and seeing the results in print, so that none of that stuff is a barrier later. This is inoculation against the inevitable future Impostor Syndrome telling you that you can’t do this, because you already did this.

To add to JeffE's answer: Yes, and it is encouraged for inexperienced researchers.

Check Eamonn Keogh's awesome slides on how to do research:

http://www.cs.ucr.edu/~eamonn/Keogh_SIGKDD09_tutorial.pdf

In particular, read from page 20: "Finding research problem":

Some people have suggested that this method can lead to incremental, boring, low-risk papers...

• Perhaps, but there are 104 papers in SIGKDD this year, they are not all going to be groundbreaking.
• Sometimes ideas that seem incremental at first blush may turn out tobe very exciting as you explore the problem.
• An early career person might eventually go on to do high risk research, after they have a “cushion” of two or three lower-risk SIGKDD papers.

Data mining is not my field, but I heard that KDD is the top-tier conference there.