FU Berlin, Fachbereich Mathematik und Informatik, Institut für Informatik

Vortrag des Informatik-Kolloquiums

Iterative Rank based Methods for Clustering

Prof. Dr. Sören Perrey, FH Gelsenkirchen 

Keywords: clustering, hierarchies, (dis)similarity, ranking procedures, statistics

Recently a new clustering algorithm was developed, useful in phylogenetic systematics and taxonomy. It derives a hierarchy from (dis)similarity data on a simple and rather natural way. It transforms a given dissimilarity by an iterative approach. Each iteration step consists of ranking the objects under consideration according to their pairwise dissimilarity and calculating the Euclidian distance of the resulting rank vectors. We investigate  alterations of this order of steps as well as  substitute the Euclidian distance by standard statistical measures for series of estimates. We evaluate the resulting different procedures on biological and other data sets of different structure regarding their underlying cluster systems. Thereby, potentials and limits of this kind of iterative approach become obvious.