I have a new clustering and linkage approach that I am testing. My algorithm detected 13 clusters when in reality there are three. I am using linkage to merge clusters by dissimilarity. That is, I initially compute the pairwise dissimilarity between all 13 clusters and then merge the two that are the most similar. I repeat this step until only one cluster remains and then examine the dissimilarity levels at each stage.
I have the following two lists in Mathematica:
DissData={0.0312488, 0.0524895, 0.064099, 0.0686212, 0.0961492, 0.102107, 0.103257, 0.125655, 0.133191, 0.162344, 0.327446, 0.546044};
WhichOneGotMerged={{5, 6}, {10, 11}, {2, 4}, {6, 9}, {5, 8}, {6, 8}, {2, 3}, {4, 5}, {2, 4}, {2, 4}, {1, 3}, {1, 2}};
The DissData list at each position contains the dissimilarity between the corresponding clusters in that position in WhichOneGotMerged. That is, at step one, clusters 5 and 6 were merged as they were the most similar with a dissimilarity level of only 0.0312. Their new cluster is relabeled cluster 5 then cluster 7 becomes labeled cluster 6 and so on until cluster 13 becomes cluster 12. Then it is repeated and cluster 10 is merged with cluster 11. And so on...
I want to plot a dendrogram, preferably using the existing Dendrogram function which will outline this process and will display the original labels of each cluster. Simply taking Dendrogram[DissData] won't work as it won't keep the labels.
Any ideas as to a direction to take this would be very appreciated. Cheers S.E.!
