Permutation Clustering using the Proximity Matrix

Clustering is fundamental to extracting knowledge from data and is one of the front line attacks. It is classification without comparing to known classes. There are many clustering algorithms. This paper is a treatise on the validation of clustering through visualization of the re-ordered proximity matrix. The paper also proposes a method for extracting clusters automatically from the re-ordered proximity matrix whose density graph representation shows the clusters visually. The method does not at any stage require the specification of the number of clusters. Through simulations and comparisons the method is shown to be quite effective.