Exploration on the commonality of hierarchical clustering algorithms
Yonghui Chen, Alan Sprague · 2004
Many clustering algorithms have been devised. The published account of each algorithm emphasizes how it is different from previous algorithms. However, similarities among agglomerative hierarchical algorithms are greater than commonly supposed. For example, several algorithms perform merging by the single link (SLINK, OPTICS) and some algorithms perform merging by the edge cut criterion (CHAMELEON, ROCK). Some algorithms use the square of the original adjacency matrix (OPTICS, ROCK). Our goals are to compose a not very long list of methods used by the various algorithms; to locate each algorithm in this space of methods; and to devise new algorithms that improve upon the previous methods.