Co-occurence network visualization using for data processing and layout and Gefx.js for producing interactive, online graph in simple html5 web page.

This HOW TO is aimed at producing a basic, undirected network graph of connected entities, represented by weighted nodes.

See an example of a network visualization produced with this workflow.

R packages igraph and rgefx fell short in our tests, with clear deficiencies in the application of standard network graph algorithms and output. Gephi (the Java-based application) and Gefx.js (browser-based viz bundle for Gephi file format) are the quicker and more accessible choice.

Prepare .csv data:

a. Edges (or relations) file
Each id number must match an entity in the nodes file. Ids can appear multiple times (e.g., Librarian A is paired with Wells library; and Librarian A is paired with Moving Image Archive).

Create file with names connected to IDs.

Human Readable file for development

18,Colin Allen,InPhO,108   
58,Jaimie Murdock,InPhO,108  
27,Bernie Frischer,Informatics,71 
28,Beth Plale,Informatics,71 
29,Matthew Brennan,Informatics,71  
19,Cassidy Sugimoto,ILS,70

Modify data for Gephi
We remove labels (text strings) in the file to be uploaded, leaving only the pair of unique ids. Note: identifying source and target is unnecesary because this is for an undirected network.


b. Nodes (or vertices) file
This file will deliver our text label information. PersonID and affID are treated the same, reduced to a simple ID.

Very important: Each entity must have a unique id that is matched in the edges file. No duplicates are allowed in nodes file. In other words, each entity can appear only once, with an Id that matches each instance the entity appears in the edges file.

1,Michael Martin
2,Douglas Parks
3,Stacie King
4,Ellen MacKay

Use to upload and process data

a. Start with relations. To upload, use File :: Open
Change “Graph Type” to Undirected.
Everything else can be left default.

b. Click Overview tab >> Data table >> Nodes

c. Import nodes/entities