HierarchicalClusterHs is a small Haskell program for clustering data based on hierarchical clusterization.
HierarchicalClusterHs currently requires data points to be given in a single-space-separated data file. For example, Test.txt:
0.9832849 0.0684199
1.0394514 -0.0356402
0.9395337 0.0884182
...
To obtain 3 clusters you can run:
HiararchicalClusterHs Test.txt Result.txt 3which creates a Result.txt file with the last column representing the cluster number of the corresponding data point:
-0.9567112 0.2772381 1
-0.9941113 0.2865241 1
-0.954816 0.3392383 1
...
The order of data points in the original file is not preserved.
If the number of clusters is not supplied as a command line argument, then it is decided automatically.
HiararchicalClusterHs Test.txt Result.txtAlthough HierarchicalClusterHs currently supports only numerical data and the Euclidean distance, the library that comes with it allows supplying other data types with other distance functions.
If you supply HierarchicalClusterHs a data file with n data points, you can ask from 1 to n different clusters to be found. If you ask for 3 clusters, HiararchicalClusterHs starts with n clusters and at each stage it reduces the number of clusters as n-1, n-2, ... until 3. The number of clusters is reduced by joining two clusters that are closest to each other. The distance between two clusters is currently set as the minimum distance between two data points which are respectively obtained from the two clusters. In certain applications this may result in undesired behavior.