Cluster

John Coene

2018-11-18

Since v3.0.0 you can color the nodes by cluster. Simply use sg_cluster with your igraph clustering algorithm of choice.

nodes <- sg_make_nodes() 
edges <- sg_make_edges(nodes, 17)

sigmajs() %>% 
  sg_nodes(nodes, id, size) %>% 
  sg_edges(edges, id, source, target) %>% 
  sg_layout() %>% 
  sg_cluster()

Clustering is brilliant, in part, because (most) are bottom-up approaches; the number of cluster is not pre-defined. However, this might make the above approach somewhat awkward as the node colors should be discrete. In an interactive environment sg_cluster prints the number of clusters identified to the console, so you could then set the exact number of colors needed then re-run the graph.

Perhaps a better approach is to use the helper function sg_get_cluster to return nodes with an added column, the group they belong to. The helper also lets you set (or re-set) the node colors according to the group they belong to.

Let’s take a look at an example

set.seed(19880525) # for reproducability

# make a slightly larger graph
nodes <- sg_make_nodes(50)
edges <- sg_make_edges(nodes, 100)

# use the helper
# first run showed 11 clusters
nodes <- sg_get_cluster(
  nodes, edges, colors = c(
    "#331A00", "#663000",
    "#996136", "#CC9B7A",
    "#D9AF98", "#F2DACE",
    "#CCFDFF", "#99F8FF",
    "#66F0FF", "#33E4FF",
    "#00AACC")
  )

sigmajs() %>% 
  sg_nodes(nodes, id, size, color) %>% 
  sg_edges(edges, id, source, target) %>% 
  sg_layout() %>% 
  sg_neighbours()