I think that in a week or two I may have a problem like this. I will have a list of "concepts" and for every pair of concepts a measure of how often they are associated.
I would like to draw a diagram/map showing this association graphically in a web page. This problem sounds very much like one from cladistics
, but not being of that profession myself I would not know.
My proposed solution (ignoring edge cases and ambiguities)
- Find the pair with the highest association metric, draw them in the middle of the graph joined by a line.
- Those two nodes form the edge of the graph.
- For each node n in the edge of the graph, find the most highly associated two nodes not already taken and join them to n. They are now added to the edge and n removed from the edge.
- Go back to step 3.
- Is this algorithm (or variant) implemented as a perl module anywhere?
- If not any general thoughts?
- Are any you of you cladists?
I had another look on CPAN after looking at the word "cladogram" and came up with Bio::Phylo
. Still it's rather a lot for me to take in.