Graphical relationship visualisation
April 29, 2016 6:45 AM   Subscribe

What tools would you recommend for creating a dynamic relationship visualisation? Thinking along the lines of They Rule.

I'm trying to map all regional organisations which have something to do with "culture" in western Sweden. They can be galleries, grant bodies, municipal institutions, workshops, schools, venues, etc. I'd like to create a database over all of these, and allow others to see the relationships between them (who get's money from where, what people sit on different boards, which are cooperating, and so on).

If you take a look at They Rule, it's possible for users to choose which companies to display, as well as showing board members - so every node can become the central point and one can easier see relationships. This is also a good way to manage the abundance of data, which in a static relationship diagram can be overwhelming.

So my question is two-fold:
Do you have suggestions for good examples of previous work which could help me to think more clearly about how to categorise these things?
Do you have any suggestion for software - online or off - which could be used for this?

I don't have the data yet, so there's I have no requirements for compatibility.
posted by monocultured to Computers & Internet (7 answers total) 7 users marked this as a favorite
For visualizing networks of relationships (which are confusingly called graphs by computer science people, but aren't the kind of graphs with an x-axis and y-axis), one open source option is Gephi.

The image at the top of this article is an example.

In your example, the nodes would be the various institutions, and the lines connecting them would correspond to flow of money, or shared board members, or whatever.
posted by caek at 7:03 AM on April 29, 2016 [1 favorite]

First off it sounds to me like you want to do nodal analysis and display of data. In my mind, that means the data is really either a giantime data dump from either a SQL or noSQL database.

If it is in SQL and all you need is to get people an interface to the data quickly, personally I like the visualizations out of Tableau, plus it is business community standard at this point, so there are a lot of resources for development as well as a familiarity of the UI. On the NoSQL front, personally I like to use SlamData to SQL-ify my NoSQL dB and then pass it into tableau again...

If I need to display nodes I like to work in Gephi, but while it looks cool it is way harder to necessarily hand someone the reigns that isn't in the tool.

On edit:
Gah, beaten to Gephi! :) seriously it is a cool tool.
posted by Nanukthedog at 7:06 AM on April 29, 2016

Response by poster: Will take a look at Gephi - seems it's a steep learning curve, but I'm not one to argue with the wisdom of Hive.

FWIW, I don't have any data collected right now, so will start from scratch gathering it. At this point I have some categories, but will have to think about how I'll end up using the analysis.

I'll go take a look at the examples and poke Gephi with a stick and see what comes of it. Thanks!

Still open for other suggestion as well as examples though!
posted by monocultured at 8:05 AM on April 29, 2016

A much much more lowbrow way to do this would be with Scapple, from the makers of Scrivener. No integral data mapping-- it is just a kind of digital whiteboard where you can connect balloons with text or pictures. But it has a free trial, and it might be useful for proto-typing or at least getting a feel for what you are eventually going to end up with.
posted by seasparrow at 8:38 AM on April 29, 2016 [1 favorite]

You're describing a multilayer graph-- this is a very technical review of the subject. You need data that looks like an adjacency matrix, or matrices: square matrix where each node has a row and a column. The data in each cell is representative of what's shared between those nodes-- the "edge" between those nodes. Maybe you have one adjacency matrix that represents the number of people who sit on boards of both organizations, and another where an edge exists if the two organizations have cooperated on some project.

Whether you can visualize those layers independently or not depends on the sparsity of the graph. If you only have a handful of connections per node in each layer, you could end up with completely disconnected components of the graph. Looking at all layers of a graph together gets complicated fast.

Check out the Reddit World Map by Randal Olson-- that's a single layer, and it results in "MLP Island". That's what I mean by disconnected components.

Scattered thoughts; sorry about that. I think you're working with smaller data than Olson (or me, for that matter). But Gephi and tools like it will still need some sort of adjacency data, and Olson's blog has pointers to a lot of visualization tools.
posted by supercres at 9:26 AM on April 29, 2016

For software, you might also try VUE. Free, open source, lots of options, easy to import csv files etc.
posted by blu_stocking at 11:16 AM on May 1, 2016

Response by poster: Thanks for the suggestions all!
posted by monocultured at 2:48 AM on May 10, 2016

« Older Bra sized long-line bikini top for a 40H   |   Late-twenties guy with no friends? Newer »
This thread is closed to new comments.