Hierarchical Clustering

An implementation of hierarchical clustering.

How to use

Specifying attribute information (<attribute file>), data file (<data file>) and linkage method (<linkage>), run the following command.

ruby clustering.rb  -a <attribute file> -i <data file> -l <linkage> -g <ignored attributes>

where <ignore attributes> is an array of attributes ignored in clustering. The clustering results are saved in a_dir/ directory.

We have 3 linkage methods.

• single_linkage
• complete_linkage
• average_linkage

Next, we obtain a PGF/TikZ script of the dendrogram from the cophenetic matrix (a_dir/cophenetic).

ruby coph_to_tikz.rb -c a_dir/cophenetic 

A default output is dendro_tikz. Copy the output in a figure environment of an LaTeX file. We need \usepackage{tikz} in the preamble of the LaTeX file.

For iris.data, we set

<attribute file> := datasets/iris.attr
<data file> := datasets/iris.data
<ignored attributes> := class

Example

Run, the following.

ruby clustering.rb  -a datasets/iris.attr -i datasets/iris.data -l average_linkage -g class
ruby coph_to_tikz.rb -c a_dir/cophenetic

We obtain a dendrogram.