| 
  • If you are citizen of an European Union member nation, you may not use this service unless you are at least 16 years old.

  • Dokkio Sidebar (from the makers of PBworks) is a Chrome extension that eliminates the need for endless browser tabs. You can search all your online stuff without any extra effort. And Sidebar was #1 on Product Hunt! Check out what people are saying by clicking here.

View
 

Adapted version of Par Martin Grandjean's Gephi Tutorial

This version was saved 6 years, 5 months ago View current version     Page history
Saved by Alan Liu
on February 26, 2016 at 3:19:41 pm
 

viated, slightly edited version of the original version of Par Martin Grandjean's "Introduction to Network Visualization with GEPHI" (2013), published under a CC BY 3.0 CH license. Some instructions have been revised because of changed interface features in the newer version of Gephi (v. 0.9.1).

 

 

Note: Grandjean has since published a new version of his tutorial (2015). For the purposes of an initial workshop, however, the data set (nodes and edges .csv files) for the original tutorial is simpler to grasp conceptually. Also, the later tutorial requires additional Gephi plugins and for work with geolocation.


 

1. Short introduction to Social Network Analysis

 

GraphExample

A network consists of two components : a list of the actors composing the network, and a list of the relations (the interactions between actors). As part of a mathematical object, actors will then be called vertices (nodes, in Gephi), and relations will be denoted as tiles (edges, in Gephi).

4 types of centrality measures (Claudio Rocchini, Wikimedia)

At the left, you can observe a very simple social graph, with both lists made explicit. Two attributes are attached to the nodes : a label (his or her “name”) and a numeric attribute (akin to the sex of people here, for example). In the edge list, “Source” and “Target” entries refer to the nodes’ numeric identifiers (Id). In our example, the “sex” attribute determines the color of the nodes. The size of a node depends on the value of its “degree centrality” (its number of connexions). The centrality measures are essential metrics to analyze the position of an actor in a network. They come in many variations, as shown at right (A = Degree centrality, number of connexions ; B = Closeness centrality, closeness to the entire network ; C = Betweenness centrality, bridges nodes ; D = Eigenvector centrality, connexion to well-connected nodes).

4 types of centrality measures
(Claudio Rocchini, Wikimedia)

 


 


3. Downloading the data set for this tutorial

 

Download both the following CSV files. Be sure to save them with a .csv file extension. (If you need to, you can paste their contents into a plain-text file with a .txt extension, then rename the file with a .csv extension).

 

 

Note: "CSV" (standing for "comma separated values") refers to a family of plain-text format data files in which the values are separated by a a comma, semi-colon, tab, or some other separator. Grandjean's csv files use a semi-colon separator. CSV files are a common way to store data sets because they can be easily migrated into other formats. For example, one can open them in Excel as a spreadsheet (and one can save spreadsheets in CSV format).

 

DatasetThe data consist of a random selection of Twitter users and their “followings” relations. The “Nodes” file contains the identifiers of each nodes, their label, a sex attribute and a random value that will be usefull to play with visualization tools hereafter. The “Edges” file contains a list of identifiers couples showing who follows who.

 

 


 


4. Importing the data into GEPHI

 

GephiDataLab

GephiImport

Start Gephi on your computer and create a “new project” in the start menu. In the Data Laboratory, click on “Import Spreadsheet” to open the import window and import your “nodes” file.

ImportNodes1ImportNodes2

Nodes

Specify that the separation between your data is expressed by a semicolon and do not forget to inform Gephi that the data you import is related to nodes, as demonstrated in this example (left). Then press “next” and fill the import settings form as proposed (right).

 

 

 

 

 

 

ImportEdges1Edges

ImportEdges2

Follow the same procedure as for the nodes, but with the “edges” file downloaded above and by filling the forms in the following manner: specify the semicolon and inform Gephi that this time you import the edges. Fill in the last fields, and uncheck “create missing nodes”, because you’ve already imported them.

 


 


5. Visualization!

 

The action now takes place on the overview panel (chosen through the tab below).

GephiOverview

Clicking "overview" produces an overview of the graph spatialized randomly that is completely unreadable (below).

RadomLayout

To make the graph readable, we will adjust node sizes, apply a spatialization layout algorithm (to separate out nodes into meaningful clusters), and then use colors and labeling to make attributes visible:

 

Adjust Node Sizes

 

Setting node size in Gephi

  • In the "Appearance" panel at left of the Gephi workspace, select "Nodes" (#1 in example diagram).
  • Then click the icon indicating size adjustment (#2).
  • At that point, it’s possible to click on the “Spline” blue link (#4) if you wish to edit the shape of the spline. You can use one of the pre-set templates in the spline choices if you wish. (An explanation of "spline") (Be aware that linearly double the radius of the nodes is more than double the area because of the power function).


 

 

Spatialization (Layout in Space)

 

FR2

FR1Spatial visualization is the biggest part of what you can do in Gephi! While it is possible to play (and lose yourself) with various visualization capabilities, I propose a method appropriate to the data set for this tuturial. 

 

In the "Layout" panel at the left of the Gephi workspace, start by choosing the Fruchterman Reingold layout algorithm, use the same values ​​as in this model (10000 10; 10). Then click on "Run" and let Gephi begin adjusting the position of the nodes. (This visualization distributes nodes according to a physical metaphor of attraction-repulsion, as if the nodes were magnets). You’re already able to distinguish communities (more densely connected parts of the network). Let the function run until the graph is stabilized. You can scroll your mouse to zoom in and out of the graph. Then you can use the little blue magnifying glass (at the bottom left of the frame of the central graph panel in the Gephi workspace) to re-center the graph.

 

FA1FA2Then, in a second stage of spatialization, use the Force Atlas 2 layout algorithm to disperse groups of nodes further, providing more space around larger nodes. (Be careful, the parameters you enter can significantly alter the final appearance.) In the parameter options for the Force Atlas 2 layout, check “prevent overlap” and change “Scaling” to 10).  Then click on "Run." Let the function run until the graph is mostly stabilized.


 

Set Node Colors

 

Setting node color in GephiNodescolor2

  • In the "Appearance" panel at the left of the Gephi workspace (with "Nodes" still selected), click on the icon for setting colors (#1 in diagram at left).
  • Then click on "Attribute." Because the nodes in this tutorial data set have attributes (gender), you can color them regarding their “sex” attribute (denoted by a value of 1 or 2 in the csv file) or simply with their degree centrality.
  • To color by attribute, click on the drop down menu (#3 in diagram) and choose the attribute you wish to differentiate by color.


 

Label the Nodes

 

Set node labels in GephiTo display labels for the nodes based on their attributes, do the following:

 

  • At the bottom right of the graph display in the Gephi workspace, click on the icon for "label text settings" (#1 in diagram at left).
  • A little panel will open up on which you can choose the attribute(s) you wish to see displayed as labels, e.g., "sex" (#2 in diagram).
  • , you find a little sign which allows you to develop a new panel. In “Label“, check “Nodes” to add their labels to your nodes and set their font/color/size… If wanted, you can click on the “Configure” link to set the data you want to get displayed.

For privacy reasons, the names are all displayed as “Names” in this dataset. No doubt you understand. Moreover, as the data set has been built (and not collected in a rational manner), the data can not be subject to interpretation.

PreviewFinal details

Go to “preview” for trimming the final details. Unlike during previous stages, changing settings in this menu is reversible, and do not affect the structure of the graph. In the this screenshot, you will find a suggestion of settings for a good rendering. Be aware that due to its large size, the graph may take a few seconds to update after each change (click on “refresh” to apply the changes).

At the bottom of this preview column, you find an export link. Note that exporting in .png produces figure with a poor resolution. You may want to opt for .svg or .pdf, which have the advantage of being modifiable by your own image/drawing software (I recommend the open source program inkscape for manipulating .svg files).

 


6. Other features

StatisticsThe visualization is only one step, network analysis often needs other mathematical means to provide the researcher with a satisfactory result. Feel free to explore the “Statistics” menu (right), for example by playing with degree measures, density, path length, modularity.

A network contains internal subdivisions called communities. There are methods that permit to highlight these communities, which depend on the comparison of the densities of edges within a group, and from the group towards the rest of the network. More here!

In the right column of the “overview” page, click on Statistics/Modularity/Run to display the modularity window. Choose a resolution (between 0.1 and 2), click OK and close it.

Modularity1Modularity2The next step takes place in the Partition menu situated in the left column. Select “Nodes” and “Modularity Class” (rolling menu). You will be then able to modify the colors attributed to the detected communities by clicking on them.

Do not hesitate to repeat this operation with many “Resolutions” ! If you decide to do so, you must deselect and reselect “Modularity Class” in the left column, and refresh color calculation.

 

 

 

 

Comments (0)

You don't have permission to comment on this page.