add alternative approach from 2013-01
[synclust.git] / R / clustering.R
CommitLineData
15d1825d
BA
1#main function (choice between kmeans and hierarchical clustering)
2getClusters = function(distances, method, K)
3{
4 clusts = c()
5 if (method=="KM")
6 {
7 nstart = 10 #number of kmeans random restarts
8 maxiter = 100 #maximum iterations count in each km run
9 clusts = .Call("kmeansWithDistances", distances, K, nstart, maxiter)
10 }
11 else if (method=="HC")
12 {
13 #simple hierarchical clustering using ECT distances
14 hct = hclust(as.dist(distances),method="ward.D")
15 clusts = cutree(hct, K)
16 }
17 return (clusts)
18}
19
20# renumbering step (post-processing after clustering)
21reordering = function(clusts)
22{
23 newCl = clusts
24 maxInd = max(clusts)
25 counter = 1
26 for (i in 1:maxInd)
27 {
28 if (sum(clusts == i) > 0)
29 {
30 newCl[clusts == i] = counter
31 counter = counter + 1
32 }
33 }
34 return (newCl)
35}