MATLAB: Question about the resulting dendrograms in hierarchical clustering

clusteringhierarchical clusteringunsupervised

I have a question about the resulting dendrogram in hierarchical clustering.

I have 64 items that need to clustered by hierarchical clustering. Following the documentation in Matlab, I computed the pdist (because each item is a vector of size 6); then I computed for the linkage.

And the resulting dendrogram is shown here:

Question:

1. I am supposed to have 64 items, where are the rest? I am only seeing a few of them? I tried zooming out but I didn't see them.

Best Answer

Dendrogram defaults to 30 leaf nodes to draw. To draw more, specify # of leaf nodes to draw.

Y = linkage(rand(64, 3));
dendrogram(Y, size(Y,1)+1)

https://www.mathworks.com/help/stats/dendrogram.html

Related Solutions

MATLAB: Custom Dendrogram Ordering Implementation

Apparently, none of the available functionalities of optimalleaforder does the work. So here it is a function that runs through the rows of the tree matrix generated by linkage and produces your output.

function c = binOrd(tree)
      p = size(tree,1);
      c = zeros(1,p+1);
      k = 0;
      recBin(p);
      function recBin(r)
          x = tree(r,1:2)-p-1;
          if x(1)<=0
              c(k+1) = tree(r,1);
              k = k+1;
          else
              recBin(x(1));
          end
          if x(2)<=0
              c(k+1) = tree(r,2);
              k = k+1;
          else
              recBin(x(2));
          end
      end
end

To get your desired rightmost binary ordering, run dendogram with 'Reorder' set as the output of the binOrd function:

tree = linkage(D,'average');
myBinOrder = binOrd(tree);
dendrogram(tree,'Reorder',myBinOrder);

MATLAB: Assigning a new data point to a cluster in hierarchical clustering

If you know your maxclust is 2, then just use 2-means. Hierarchical clustering and linkage will give you an insight about the number of natural clusters in your data. When you found that k' (e.g. using 'cutoff') based on hierarchical clustering, you can use k'-mean.

Best Answer

Related Solutions

MATLAB: Custom Dendrogram Ordering Implementation

MATLAB: Assigning a new data point to a cluster in hierarchical clustering

Related Question