The Witch’s Brew: VOSviewer and Graphia
Harry Potter series is great for learning a lot of things, not necessarily wizardry. In “The Witch’s Brew” series I learn various softwares for analyzing networks. I test them before proposing them to my students. Some are just what I am familiar with, like Gephi. Others, are what I am trying to learn, like Python libraries. In this post I present two more, which were originally created to study different kinds of networks, but can be used to accommodate any kind — with a little patience.
Best read listening to Faun’s ‘Walpurginsnacht’
There are more open source software solutions around the corner, though I am not that familiar with them. So please read the documentation to VOSviewer and Graphia, and the nice introduction to Graphia by Dr Veronica Espinoza.
VOSviewer (the online version)
VOSviewer is a software tool for constructing and visualizing bibliometric networks. These networks may for instance include journals, researchers, or individual publications, and they can be constructed based on citation, bibliographic coupling, co-citation, or co-authorship relations. VOSviewer also offers text mining functionality that can be used to construct and visualize co-occurrence networks of important terms extracted from a body of scientific literature.
Of course, networks are networks, and all sorts may be visualized in VOSviewer, as shown in the next two images.
I used the the online version at the https://nocodefunctions.com created by Clement Levallois to import the gexf file, since the standalone version does not import gexf files. The fist thing I got was once again sort of a ‘hairball.’
There are several options allowing for changing the layout until the user gets a satisfying visualization (satisfying here means that it allows for explanation).
One thing with VOSviewer online is that it computes statistics behind the scenes, so the user has to bother only with visualization. This is good for a newbie in networks, but not enough for analysis: how could you compare networks if you don’t have access to statistics like Average Degree, Diameter or Density, Components or Modularity Classes? Though, with some experimentation, network graphs look interesting and inspiring.
Graphia
On the other hand, Graphia is a powerful open source visual analytics application developed to aid the interpretation of large and complex datasets. Graphia can create and visualize graphs from tables of numeric data and display the resulting structures. It can also be used to visualize and analyze any data that is already in the form of a graph.
Graphia does not accept gexf files as import. So, I exported the data in graphml from Gephi and opened the file in Graphia.app.
While a lot of metrics can be applied, here called Transformations, not much is easily shown to the user. Even so, Graphia can work with really big networks, and with a bit of extra training it may produce amazing images. In our case, a lot of ingredients are known but their use remains unknown, so they form a circle of blue nodes around the main network. Had they been removed (as in the case of VOSviewer) perhaps a clearer version of the main component would appear. Also, there are several ingredients applicable to only one recipe, some even in a relationship one-to-one. They are at the periphery as well.
Graphia supports different modularity algorithms than Gephi, Leiden is one of them (supported by Gephi as a plugin), and MCL the other.
What is amazing is that you get a 3D representation of the network.
Really cool if you need to understand the network structure beyond the confinements of two dimensions.
Coming next: Take it or leave it solutions for visualizing networks. Stay tuned.
