First post :)
I'm going through the 'Analysing social networks with NodeXML' book but am struggling with the change of terminology and functionality from clusters to groups.
The book (esp chapter 7) talks about clustering but in my version of nodexl this is now groups. There's a lot of discussion about making use of the 'Find Clusters' feature (esp wrt Les Miserables data) but an alternative option doesn't seem to be available
within the Group menu.
Can anyone offer advice?
Feb 3, 2012 at 1:21 PM
This may help: bit.ly/g5zcnI NodeXL: Clusters, components and groups – Creating & managing collections of vertices.
See also: "Creating Groups" in NodeXL help system. Group vertices by attribute, component, algorithm, or (coming soon) "motif".
NodeXL allows collections of vertices in a network to be gathered together into a “Group”. Groups have several properties:
- groups can be selected
- vertices in selected groups can be operated on as a set
- groups can be collapsed or expanded
- network metrics can be calculated for each group
- groups can be plotted within bounded regions
NodeXL supports creating clusters or groups of vertices in several ways: by attribute or manually, by component, or algorithmically.
Group menu commands are located in the Groups Menu in the NodeXL>Analysis Menu section.
Group menu commands include:
Group by Vertex Attribute
Users can also assign vertices to groups based on any attribute in the vertex worksheet.
The NodeXL>Analysis>Groups>Group by Vertex Attribute allows groups of vertices to be defined
by any attribute.
These attributes can be numeric, or categorical:
Groups can also be authored manually. A group is created whenever a new row is populated in the Groups worksheet. A vertex is assigned to a group when it is named with its group in the Group Vertices worksheet.
Find Connected Components:
Each component can be assigned to its own group using the NodeXL>Analysis>Groups>Find Connected Components option.
Find Clusters – Automated Group Assignment Algorithms:
NodeXL exposes three of the clustering algorithms from the Stanford
Network Analysis Platform (SNAP) (http://snap.stanford.edu) library for calculating network metrics from graphs. Working with SNAP author and Stanford Computer
Science Professor Jure Leskovec, the NodeXL team integrated three clustering algorithms which can be selected from Cluster>Options:
the Wakita and Tsurumi “Finding
Community Structure in Mega-scale SocialNetworks“ algorithm, the Girvan-Newman or Clauset-Newman-Moore algorithm.
When these Group findings algorithms are run each vertex is assigned to one of a set of groups based on its decision rules. In general, these algorithms try to place collectons of densely connected vertices into separate groups or clusters.