0
$\begingroup$

I'm looking for a distance metric to compute how close certain paths taken by people navigating throughout a city are to a set of 'correct' routes.

I have path recordings for some 'correct' routes throughout a city. I also have path recordings for users navigating the same city. Each recording consists of rows of sequential X and Y coordinates. There is some variability in when X and Y locations were sampled, so there is not a one-to-one correspondence between rows in different recordings. Recordings are also of variable length.

Any suggestions on how best to estimate how close the routes taken by people were to my pre-recorded correct routes? I'm not fussed about time differences, so taking the exact same route but faster would not be distant.

$\endgroup$
  • $\begingroup$ The answer ought to depend on what use you will make of this "closeness:" what is the purpose of your analysis and how do you intend to interpret these values? $\endgroup$ – whuber Aug 12 at 12:26
0
$\begingroup$

I don't know of one off the shelf but you could come up with one. For example, the distance between two vertices on a graph is the length of the shortest path connecting the two vertices. Then your distance metric between paths could be the sum of the distances between vertices in path A and the closest vertex in path B.

If you are not working with graph structured data explicitly (though you should think about it since moving along roads/paths in a city is not the same as moving freely around $\mathbb{R}^2$) then you can do a similar trick integrating some measure of distance between points for every point on the path.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.