It can certainly be thought of as a math problem. I haven't read much about it or thought much about it and for now I'm refraining from looking at any kind of algorithm or solution (not that I plan on working on it or anything right now -- maybe just thinking about it). But a shortcut to discerning the complexity of the problem (for humans at least) is by recasting it as a graph theory problem, in which you must find a Hamiltonian cycle in a weighted graph. Unlike the more amenable Eulerian paths, Hamiltonian paths do not have a complete theory associated with them. This alone shows the complexity of the problem. The sufficient conditions discovered for Hamiltonian paths/cycles are very weak, e.g.: A graph with v vertices has a Hamiltonian cycle if each vertex has degree of at least v/2. This condition is weak because it forces such an uncommon situation.
Anyways, kinda cool that you brought it up because I've been doing a lot of graph theory lately.