Cluster Dialog and Algorithm

 

TouchGraph Navigator identifies clusters are determined based on Edge Betweeness Centrality (external pdf)

 

Nodes within clusters tend to be close together, while nodes in separate clusters are further apart.  TouchGraph's cluster algorithm identifies long edges which are likely to connect separate clusters and ignores them. As more and more edges are ignored graph is split out into separate connected components which form the clusters.

 

 

The minimum number of clusters that the graph can be split into is the number of connected components that are in the graph to begin with.

 

One can use the # Clusters Displayed slider and spinner to adjust the number of clusters that can be formed by ignoring edges, as specified above.

 

The cluster computation algorithm used by TouchGraph is fast and non-deterministic.  It may create slightly different results each time.  The Recompute Clusters button allows you to try the algorithm again to get a different grouping.  Only nodes that are on the border between clusters are usually affected.

 

Clusters are assigned different colors in the graph.  Sometimes two adjacent clusters get assigned the same color.  In such cases it may be useful to use the randomize option in the Node Appearance - Colors dialog to change the cluster colors so that splits between clusters can be revealed.

 

 

 

The Ok button closes the cluster dialog.