cluster_redwalk: Berenhaut Community Detection on Networks
In barrpet/gclust: Network Community Detection

Description Usage Arguments Value Examples

Performs community detection on networks using average shortest path distance of vertex neighbors as the set of dissimilarities for heirarcal clustering with "average" as the agglomeration method. If shortest paths are supplied, this function is ~50% faster. It is also best to use a symmetric "matrix" rather than a "dist" object if performance is critical. Time complexity is O(n^2 log n) and space complexity is O(n^2). This function returns an hclust object rather than an igraph communities object because the algorithm does not attempt to use modularity use or maximize modularity (although the resulting partitioning results in a maximum modularity that is competitive with the top community detection algorithms).

1	cluster_redwalk(graph, nodes = V(graph), short_paths = NULL)

`graph`	an igraph object.
`nodes`	a subset of the nodes in the graph to cluster, defaults to all nodes (full community detection)
`short_paths`	either a \|V\|x\|V\| symmetric dissimilarity matrix of the vertices in `graph` or a `dist` object from the stats package. If `NULL` (default), the shortest paths will be calculated. Note: much faster if `short_paths` is provided and much faster if `short_paths` is of class "matrix" rather than "dist".

cluster_redwalk returns a hclust object from the stats package.

require(stats, quietly = TRUE)
require(Rcpp, quietly = TRUE)
cb <- cluster_redwalk(karate)
## cb is an hclust object
plot(cb)
## membership for different community sizes (k)
cutree(cb, k = 2)
cutree(cb, k = 3)

## using precomputed shortest paths
sp <- shortest_path_lengths(dolphins)
cbd <- cluster_redwalk(dolphins, short_paths = sp)
plot(cbd)
cutree(cbd, k = 2)
cutree(cbd, k = 4)