Visualizations : Test d

Bernice again

Mercan Topkara: I just thought that if I knew these people (if the names meant something to me) I would have arranged the graph differently. Those who I know more would be in the center, or those whom I met at the same time would be closer for me as a memory clue.

comment

First: minimize cumulative link length, while maintaining a minimum length per link, which results in nodes with a lot of links placed centrally
Second: minimize link crossings, to reduce visual clutter and to reduce overlap between the networks around individual nodes.

2 clusters connected with only 3 connections

I organization of this visualization based on increasing the white space and trying to reduce “tangle” by cutting down on the number of edge intersections. With the exception of “Eric” and “Amanda”, I think it works well.

I’m simply trying to simplify crossovers and have as few overlapping network links. I’d like to be able to use this software in my work…

Having the graph as planar as possible seems helpful. Allowing a different edge color for edges that cross others would help.

I tried to make as few lines cross as possible. I started to notice that there were two highly connected network that were connected by a few common members. The most frustrating aspect was making space inside each network for the sub networks to have room.

Recently, I have been reading the companion to Mark Lombardi’s art work, Mark Lombardi: Global Networks. I noticed how his connector lines were all mostly curved. It made me rethink using straight and right angle connectors in my flow diagrams because the arc of his lines create a “trajectory” that is easy to follow with the eye even when crossing several other lines.

as compact as I could get it

I like it. It has nice shape and relations makes sense.

Nodes with more links should probably be moved first. One factor is how much overlap there may be between links.
Some fairly obvious clusters.

ugh

two groups with few ties

In trying to minimize crossing, I usually try to get the three-path-out nodes triangulated within their destinations first, then try to massage the heavier nodes to minimize the remaining crosses. I also try to get line-crossings to happen relatively close to the midpoint of each line, and to make sure that a given path doesn’t have to “squeeze” between two points that are very close.

I rearragned to make sure that there were fewer crossed relationships. This naturally led me to see that people clumped and that there were only a few that had general cross over connections.

cool

I guess that the way I did it was to drag nodes around until the connections didn’t overlap, weren’t too long (although they aren’t really minimum length) and the nodes are close to lying on a rectangular grid. At some point it began to make sense…