last modified: 2021-09-25

gephi logo 2010 transparent

Description of the project

This project is for complete beginners to Gephi. It supposes you have Gephi installed and running on your computer. That is all.

When finishing this tutorial, you should be able to:

  • be familiar with the vocabulary to discuss networks

  • download a network file for this exercise

  • description of the file / the network

  • open a network file

  • read the report after opening a file

  • show the labels of the nodes

  • layout the network

  • visualize attributes of the network

  • prettify the network for enhanced readability

  • compute the centrality of the nodes in the network

  • visualize attributes created by Gephi

  • export a visualization as a picture or pdf

  • export a web visualization of the network

be familiar with the terminology to discuss networks

Figure 1. terminology for networks

download a network file

download this zip file and unzip it on your computer.

You should find the file miserables.gexf in it.

Save it in a folder you will remember (or create a folder specially for this small project).

description of the file / the network

This file contains a network representing "who appears next to whom" in the 19th century novel Les Misérables by Victor Hugo[1].

A link between characters A and B means they appeared on the same page or paragraph in the novel.

The file name ends with ".gexf", which just means this is a text file where the network information is stored (name of the characters, their relations, etc.), following some conventions.

open the network in Gephi

  • open Gephi. On the Welcome screen that appears, click on Open Graph File

  • find miserables.gexf on your computer and open it

gephi welcome screen open graph en
Figure 2. welcome screen

read the report after opening a file

A report window will open, giving you basic info on the network you opened:

opening file report en
Figure 3. report window

This tells you that the network comprises 74 characters, connected by 248 links.

Links are undirected, meaning that if A is connected to B, then it is the same as B connected to A.

The report also tells us the graph is not dynamic: it means there is no evolution or chronology, it won’t "move in time".

Click on OK to see the graph in Gephi.

initial view

project initial view
Figure 4. initial view when opening a graph

This is how the network appears in Gephi. Not very useful! Let’s examine what we have here.

basic view of Gephi’s interface

Figure 5. the 3 main screens in Gephi

Gephi has 3 main screens:

  1. Overview: where we can explore the graph visually

  2. Data Laboratory: provides an "Excel" table view of the data in network

  3. Preview: where we polish the visualization before exporting it as a pictue or pdf

What we see here is the Overview.

Figure 6. Filters and statistics panels in Gephi

In the Overview, the graph is shown at the center. Around it, several panels help us fine tune the visualization.

  1. "Filters", where we can hide different parts of the network under a variety of conditions

  2. "Statistics", where we can compute metrics on the network

Figure 7. Appearance and layout panels in Gephi
  1. "Appearance", where we can change colors and sizes in interesting ways

  2. "Layouts", where we can apply automated procedures to change the position of the network

Figure 8. 3 groups of icons
  1. A series of icons to add / colorize nodes and links manually, by clicking on them

  2. Options and sliders to change the size of all nodes, links, or labels

  3. More options become visible if we click on this little arrow head pointing up

showing labels of the nodes

Figure 9. showing node labels

layout ("spatialize") the network

Figure 10. selecting the force atlas 2 layout
Figure 11. changing a few parameters and launching the layout
Figure 12. result of Force Atlas 2 layout

visualize the properties of the nodes

A network consists in entities and their relations. This is what we just visualized. Yet, the properties of these entities remain invisible.

For instance: the characters in the novel "Les Misérables" are male or female. Are males more likely to be connected to males, or females? Just looking at the network in Gephi, we can’t tell.

Now, we will see how to make this property ("Gender") visible.

Figure 13. Switching the view to the data laboratory
Figure 14. We see there is a Gender attribute for each character.

We will color the nodes based on their gender. To do that, we select Gender in the Appearance panel:

Figure 15. Coloring nodes according to their gender

The result:

appearance miserables result
Figure 16. After coloring characters according to their gender

prettify the network for enhanced readability

There are a number of issues with the result we get:

  1. the network is too big or too small, it is hard to read

  2. the labels of the characters overlap

  3. the size of the labels might be too big / small

  4. the links are sometimes too large

Let’s fix these issues.

1. Enlarge or shrink the network

  • either we use the "scaling" parameter of the layout, as we have seen here.

  • or the scale is fine, it is just that we need to zoom it or out. Use the scrolling wheel of your mouse, and right click to move the network.

2. Prevent the Labels from overlapping

In the layout panel, choose "Label Adjust" or "Noverlap": these layouts will move the nodes just so that the Labels stop overlapping:

choosing a label adjust algo en
Figure 17. Noverlap or Label Adjust will help you

Don’t forget to click on "Run" to apply these layouts.

3. Changing the size of the labels

Open the bottom panel of Gephi by clicking on tiny arrow head (1). Then select "nodes" (2), then move the slider (3).

Figure 18. Adjusting label size
Figure 19. Adjusting edge thickness

computing the centrality of the nodes

1. Definitions of centrality

"Centrality" is a very good metrics to first get an idea of a network. What does centrality mean? Intuitively, we understand that a "central" node will probably sit in the middle of the network. But how to measure that "scientifically", so that we have an objective confirmation of our visual impression?

There are several ways, all equally interesting.

We can measure degree centrality. "Degree" is the technical term for "number of connections that a node has".

So, degree centrality just means that the most central node is the node which has the most connections. Simple!

Another measure is betweenness centrality. This one is more tricky.

  • First, you have to imagine what is a shortest path.

  • A path from node A to node B is a chain of nodes, the road if you will, that you have to traverse to go from A to B.

  • The shortest path from A to B is the quickest road from A to B: the path that has the smallest number of nodes between A and B.

  • A node which is on many shortest paths is "between" many nodes. And when you realize it, it is a very intuitive sense of what it means to "be central". These nodes have a high betweenness centrality.

2. Computing betweenness centrality with Gephi

Gephi computes it for you. Find "Network diameter" in the statistics panel and click "run":

Figure 20. Computing betweenness centrality

This will open a window with parameters (explained in a more advanced tutorials). Click "OK":

Figure 21. Parameters for the computation of betweenness centrality

A report window opens (also explained in a other tutorials). Close it.

Figure 22. Report after the computation of betweenness centrality

Now we can visualize this information.

visualize attributes created by Gephi

Gephi has computed for us the betweenness centrality of all nodes. This remains invisible on the network, however.

It would be interesting to, say, resize the nodes according to their centrality: the more central a node, the bigger. This would allow for a very quick visual appreciation of which nodes are the most central.

First, let’s switch to the data laboratory to see how Gephi stored the "betweenness centrality" of each node:

Figure 23. Switching the view to the data laboratory

When we ran "Network Diameter" in the statistics panel, Gephi has actually computed many kinds of centralities (not just "betweenness centrality"):

Figure 24. Different centrality measures visible in the data laboratory

To resize the nodes according to the value of their betweenness centrality, we use the Appearance panel:

make sure you select the correct options
Figure 25. Ranking node sizes by centrality
Figure 26. Selecting the minimum and maximum sizes of nodes
ranking centrality miserables 3  en
Figure 27. Result of the ranking
Figure 28. Resizing labels to reflect their node’s size
result label resizing en
Figure 29. Result of the label resizing

exporting a network as a picture

1. exporting a screenshot from the Overview (a png image)

Figure 30. Open the configuration panel for screenshots

Select the maximum value for anti-aliasing, and multiply values for width and height for higher resolution. For example, resolution x 3 is width = 3072 and height = 2304

configuration screenshot en
Figure 31. The configuration panel for screenshots

2. exporting a pdf or svg picture

Let’s switch to the preview panel:

Figure 32. Switching to the preview panel

The preview panel is dedicated to the preparation of the picture to be exported as a pdf or svg, which are "scalable": in pdf or sv, the resolution of the graph will remain perfect, even with a strong zoom. But as you see, it means the network is now looking different than in the Overview.

contrary to the Overview panel, here you need to hit the "refresh" button after each parameter change.
Figure 33. Updating the parameters

Here I change just 2 parameters: Show Labels and Font (which I reduce to size "5"), to get:

Figure 34. Result of preview

Export: just click on the button and select the file format you prefer

Figure 35. Export button

3. download the result file

download this zip file if you need to see the network in its final form.

export a network as a web visualization

You can easily create a web-based version of your network. It can be made public or private. If you choose to make it public, it will be easy to share - simply send the url.

The web version makes use of VOSviewer online, which works best for networks up to ~ 500 nodes (if your networks has more nodes, it can be hard to read on screen).

To proceed, simply:

1. export your graph from Gephi as a .gexf file

make sure you export a gexf file, not a GEPHI file!

In Gephi, do: File → Export → Graph file…​ See an illustration below:

Figure 36. Export gexf file 1
Figure 37. Export gexf file 2

2. visit nocodefunctions and convert the gexf file to a web visualization

Nocode functions is a web application that I developed. It provides useful tools to work with networks and text. It is free, does not need a registration, and is respectful of your data.

  • Go to this page on nocodefunction.

  • Choose option 1 ("convert a Gephi gexf file to VOSviewer json format")

  • If you want to share your visualization, check the box to make it public

  • Click on "Convert to VOSviewer json file"

You will be directed to a new webpage like the one below:

vosviewer result
Figure 38. An illustration of the web view of a gexf file

3. public or private web visualization? How to manage

If you have chosen "public" at the previous step you can now bookmark the web page of your network, share it, etc. Otherwise, the visualization will be deleted after an hour. This is to keep the promise that your data is secure: if you want it private, we simply don’t keep it on the servers.

What if you want to keep your visualization private, but don’t want to redo all the steps after each hour?

  • in the VOSviewer visualization, save your network as a json file by clicking on the "save" icon:

Figure 39. Saving the network to a json file from the VOSviewer online interface

Then, you can re-open this file in the web interface whenever you want:

Figure 40. Opening a network in the VOSviewer online interface

4. More options to share your visualization with VOSviewer online

This web page offers a full view on the other ways you can share a network visualization with VOSviewer online.

the end

questions and exercises

  1. Open the file miserables.gexf with a text editor (here is how to do it on a Mac, and on Windows). See how the nodes and the links are written in the file. Can you find the character Javert?

  1. Our network of Les Miserables characters was undirected. Can you think of networks which are directed?

Imagine how undirected and directed networks differ when computing centrality, for example.

  1. Force Atlas 2 is a layout which brings together connected nodes, and spreads out unconnected nodes. We might have nodes with no relation at all with other nodes (called "isolated nodes").

How will these isolated nodes move on screen?

  1. When applying the Force Atlas 2 layout, the network moves quickly, then stabilizes, and then keeps moving a bit.

Can you guess why this is happening?

  1. In the list of layouts, Force Atlas 2 is just one of many options you can choose. Try "Fruchterman Reingold" and "Yfan Hu".

These are layouts which follow the same logic as Force Atlas 2, but with slight variations. Explore how these algorithms result in similar, yet specific layouts.

  1. In this tutorial, we defined degree centrality. Can you imagine a situation when a node with the largest degree centrality will actually be in the periphery of the network? You can draw a toy network to help you figure.


1. D. E. Knuth, The Stanford GraphBase: A Platform for Combinatorial Computing, Addison-Wesley, Reading, MA (1993)