4
$\begingroup$

I'm looking for info about a particular kind of graphing technique. Basically, you have a list of elements, and data indicating how closely related the elements are to each other. On the plot, elements that are closely related are plotted close to each other, and ones that are less related are drawn farther away. I'm trying to find out what this kind of graph is called and if there are any related good algorithms for it.

Sorry if this is a bad question, but this has been driving me nuts and it's impossible to search for.

Update: So it was a network graph I was looking for after all; I found this neat freeware tool called Gehpi that pretty much serves my needs. Thanks for the help!

Improve this question
$\endgroup$
4
  • 1
    $\begingroup$ My best guess would be a network graph, see an example here. $\endgroup$
    – Andy W
    Commented Nov 26, 2012 at 19:43
  • $\begingroup$ Yeah, that's definitely in the same ballpark. I can't find much info about "network graphs" besides a few images... nothing on Wikipedia, etc. $\endgroup$
    – Andrew
    Commented Nov 26, 2012 at 19:53
  • 1
    $\begingroup$ See the wikipedia page on graph drawing. That and perusing the questions here tagged networks will hopefully be a good start (if not at least to identify if that is what you are or are not interested in). $\endgroup$
    – Andy W
    Commented Nov 26, 2012 at 20:01
  • 1
    $\begingroup$ For a discussion of some graph embedding and drawing algorithms found in software, see reference.wolfram.com/mathematica/tutorial/… and scroll down the middle starting at "Graph Drawing Algorithms". $\endgroup$
    – whuber ♦
    Commented Nov 26, 2012 at 20:56

3 Answers 3

Reset to default
4
$\begingroup$

The general problem is often called low dimensional embedding or for your particular case, 2D embedding.

Isomap, Multi-dimensional Scaling, and Locally Linear Embedding are the names of three popular algorithms for the problem.

This has a nice visualization of some stock market data with a 2D embedding.

Improve this answer
$\endgroup$
7
$\begingroup$

I think you might want MultiDimensional Scaling (MDS). This takes a matrix (or one triangle of a matrix) of "distances" and gives a 2 dimensional (or other dimensional) set of points that represents those distances as close as possible.

Improve this answer
$\endgroup$
0
$\begingroup$

In general you may refer to cluster analysis where the distance between all elements is calculated. It could be also factor analysis or principal component. You can try to google for image showing above or simply statisitcal distance analysis plot and provide a link to the object describes your question.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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