Computer Scince PhD School Selection: focus on ranking rather than the group?

[ialuronico]

I was wondering how much the University ranking is important for a PhD school selection.

I think that it could really be established a discussion about University rankings and their reliability. It seems to me that top schools are there because they are the most famous ones, even in the subject ranking.

More specifically, what is the most significant thing in the PhD school selection process, the University ranking or the specific group importance?

In particular, if I could, I would like to compare Melbourne University with Waikato one about the Machine Learning (ML) research group. The first one has significantly higher ranking than the second one, but the latter has a very strong ML group. In fact, they are still publishing new versions of their Data Mining book.

Thank you.