Seems this should be a FAQ, but I do not have the technical vocabulary for an effective search.

I have a large number of observational units (~4000) in ten groups. Group sizes vary. For each O.U. I have a small fixed number of measurements (think: 4). I want to visualize this data set graphically. I was thinking of numbering the observations and then using a line graph, but the variation is so large that the graph is unreadable. But the order of the O.U. within each group is arbitrary, so I am thinking of permuting the O.U. in such a way as to minimize the chatter, that is, the sum of distances between consecutive points. (If there were a single measurement this would be trivial: sort the observations in increasing or decreasing order within each group.)

I can formulate the problem as a (mixed) binary program with linear (or nonlinear) objective, but this brute force approach seems impractical. Cbc uses around 45 seconds on a small test case (30 O.U.) with fewer than 200 binary variables. The full set is going to have around 1 million B.V. Are there shortcuts to the formulation, to the solution, or even to prior research?