<-- Advertise here.

gephi_teaser.jpg
I do not often get to play with networks, yet I find them fascinating and full of knowledge gathering potential. The data visualization works of Moritz Stefaner, Jer Thorp, and Mark Lombardi captivate me. I am intrigued by relationships, how we perceive them, and how we can understand them.

Gephi, the "Open Graph Viz Platform", is not just for the hobbiest node nerd. It is a hardcore, professional tool for network researchers and other smarty pants. If you know the difference between Yifan Hu and Fruchterman Reingold, then you can use this tool effectively. But you do not need to be rich, as the software is completely free to download and use.

So what is Gephi capable off?

 
CRITERIA
Cost Free and open source (they take donations)
Ease of Use Moderate
Operating Systems Windows, Max OSX, Linux
Export Formats Image: SVG, PDF — Graph: CSV, GEXF, GraphML, GDF
Maturity (age of software) unknown
Customer Service None provided; there is community support via the forums.
Data Import Formats:
Oracle No
SQL Server Yes
Sybase No
DB2 No
PostgreSQL Yes
mySQL Yes
Excel No
Text No
other CSV, GEXF, GDF, GML, GraphML, Pajek NET, GraphViz DOT, UCINET DL, Tulip TPL, XGMML
Reviewer(s) Kim Rees
Date reviewed 7/10/2010

gephi_airline.jpg

Gephi interface displaying airline traffic data.

Pros
— Options! Lots and lots of options. Network layout, fonts, edge size, node size, borders, colors, labels, etc. You name it, it's probably in there.
— Grouping. Ability to group nodes by data points.
— Community-detection. Gephi can identify groups by using its modularity algorithm.
— Reduction. The ability to filter out practically anything in order to make a more coherent graph.
— Statistics! I'm sure this is really cool, but I'm not educated enough to know. If you know your Eigenvector Centrality from your Average Clustering Coefficient, then you will love this feature.
— Export. You can export your graph as SVG or PDF.

Cons
— Interface is a bit cluttered. Multiple levels of tabs and redundant buttons make it a bit confusing to navigate. This will probably wane with long term use of the tool.
— A bit buggy. The undo, cut, paste, etc. menu items never seem to work for me. Certain settings remained "on" even after I turned them "off."
— No documentation. There are a couple of user guides that are helpful, but lacking.

gephi_partition.jpg

Main interface displaying the Diseasome data.

Review
Overall, Gephi is a useful and interesting tool. It allows you to explore, create, and analyze network graphs in a fairly easy way. It offers many visualization options for any type of network and supports up to one million nodes and edges.

There are some plugins available for Gephi that extend the functionality. I used one called GeoLayout to make the Minard graph below.

The only thing keeping Gephi from being a professional grade tool is its bugginess and lack of documentation. If its development community can sustain it, Gephi should be able to improve and become a more robust software tool.

gephi_minard.jpg

Graph of Minard's data using Gephi and the GeoLayout plugin.

For the Purists
Gephi is a purist's delight. All the network graphing functionality is based on cold, hard science. If you recently read the Holten and van Wijk paper, you will like Gephi. It has loads of options but all within the confines of good taste.

For the Aestheticians
While there are no progressive or experimental ways of graphing networks with Gephi, you might still be happy to use it for its ease of exploration. It allows you to customize the look of your graph in many ways. It allows for proportional labels and label truncation, different types of edges based on if its directional or not, and everything from node color to group highlighting. I have to say, I was impressed by the level of visual customization. There are a few things that are overlooked, however, such as overlapping labels, annotation, and so forth. Since the graphs are SVG exportable, at least some of these visual problems can be rectified.

gephi_metrics.jpg

Two nerdy charts of network statistics that Gephi will output.

Please let us know your thoughts in the comments below! Thanks!

Kim Rees is a partner at Periscopic, a socially-conscious Information Visualization firm specializing in helping nonprofit organizations and like-minded companies convey important messages and elevate public awareness.

6 COMMENTS

Kim, this appears to be a great open source tool that I'll definitely explore. I'm interested in mind mapping and concept mapping to help us better represent knowledge.

Thanks for introducing me to a new community of mappers.

Fri 16 Jul 2010 at 2:53 AM

I tested it a few weeks ago, just for fun and I must say it is impressive. If I ever need to visualize some unordered data (the one that gnuplot can't manage, I mean) I will definitely use it, although it can be a little hard to use.

Ruben

Fri 16 Jul 2010 at 11:01 AM

Hi,


Thanks for this review, it's appreciable to see how the software is perceived by people outside the field of networks. I complete the missing information:
* Maturity/age of the software: 2 years for the project, 5 months for the latest series (see the Gephi download page on Launchpad)
* Customer Service: depends on what's behind the terms. Some services exists, you just need to ask: http://gephi.org/users/courses/


Gephi is currently in the alpha stage (nor beta, nor official release) and documentation getting slowly available, but it's not an easy task as we aim to provide both user and developer documentation. In a software development life cycle, it's when nerdy people want to use it despite these lacks, like someone who run a car built at 50%. But it seems very useful for a lot of people right now...


These people are mostly working in research laboratories and R&D departments ; there's also funny people, like aestheticians! ;-)
The hard science and the lot of options are what's necessary to produce meaningful representations (not just communication snapshots). But our cold software has also a very hot community behind it, so if you don't know the difference between Yifan Hu and Fruchterman Reingold, we'll be happy to teach you on the forum :)


"For the Purists" ?
I don't know. It is for people that have to explore and understand graphs. (source: Features page) I'm sorry if it currently requires to know what a graph is, but I'm sure we will fill the gap in the future by documenting more. We are aware that going from an Excel table to a graph representation is not an easy task. I repeat that our community is pleased to answer to questions like "What the Eigenvector Centrality is?", or "How nerdy charts help me to understand my data?". Members of the community are encouraged to be curious and discover both the software and the research studies around.


"For the Aestheticians" ?
Maybe. We're working on making things easier. We already face the overlapping label issue with a layout called Label Adjust, which...slightly adjusts the nodes positions. We have very few documentation, but it's in one of the two tutorials.


See you on the forum!
Seb

Wed 21 Jul 2010 at 7:25 PM

Hi Sebastien,

Thanks for the clarification on those points. I hope you didn't read this as a negative review; I really enjoyed using your software.

When I remarked about Gephi being for academics and researchers, it was out of reverance. I realize that everyone can use your tool, but that people with network analysis experience will find it particularly useful.

You are right that I should have mentioned that Gephi is alpha. I was really excited to use and review your software, so in my haste I may have missed a few points.

Kim

Fri 23 Jul 2010 at 5:59 PM

Hi Kim,

It's cool anyway to see such spontaneity! I was just troubled that you didn't took the time to contact us for the missing information, but it's clearer now.


Btw, we have to better support the data importers: everybody waits for an easier data flow, and that's in our high priority list. If someone is interested to contribute in coding a file importer or a db connector, we'll be happy to provide guidance and mentoring!

Seb

Sat 24 Jul 2010 at 6:32 AM

Great post, and interesting software.
Another interesting package for visualizing flows in particular is JFlowMap, also open source.
http://code.google.com/p/jflowmap/

Mon 26 Jul 2010 at 1:10 AM
ADD A COMMENT
Commenting has been temporarily disabled.