A black hole is a region of spacetime enclosed by an event horizon. Thus, the singularity, while a fact about black holes as far as we understand them, is not an defining feature of what black hole is.
Therefore, it makes more sense to try to calculate volume (and hence indirectly, density) of a black hole according to the extent of the event horizon rather than the singularity. It so happens that for the simplest of black hole solutions, the spherically symmetric Schwarzschild black hole, the Schwarzschild radius characterizes the size of the horizon: the Schwarzschild radius is actually $\sqrt{A/4\pi}$, where $A$ is the area of the horizon. This is actually (implicitly) by definition of the Schwarzschild radial coordinate, which directly corresponds to the area of a sphere.
As I've said in the comments, a problem of talking about 'the' volume of a black hole is that it's not actually frame-independent, so calculating volume in this way is simply a conventional choice. Again, see this question for more details.