Adaptive Clustering and Multidimensional Scaling of Large and Highdimensional Data Sets

Abstract

We describe a algorithm for exploratory data analysiswhich combines the adaptive c-means clustering and themultidimensional scaling procedure (ACMMDS). ACMMDS is analgorithm for the online visualization of clusteringprocesses and may be considered as a alternative approachto Kohonen's self organizing feature (SOM). Whereas SOM isa heuristic neural network algorithm, ACMMDS is derivedfrom multivariate statistic algorithms. The possibleimplications of ACMMDS are illustrated through twodifferent data sets.