clustering objects on subsets of attributes.pdf

Clustering Objects on Subsets of Attributes Jerome H. Friedman Jacqueline J. Meulmany March 14, 2004 Abstract A new procedure is proposed for clustering attributeñvalue data. When used in con- junction with conventional distance based clustering algorithms this procedure encourages those algorithms to automatically detect subgroups of objects that preferentially cluster on subsets of the attribute variables rather than on all of them simultaneously. The relevant attribute subsets for each individual cluster can be di§erent and partially (or completely) overlap with those of other clusters. Enhancements for increasing sensitivity for detecting especially low cardinality groups clustering on a small subset of variables are discussed. Applications in di§erent domains, including gene expression arrays, are presented. Keywords or phrases: distance based clustering, inverse exponential distance, clustering on variable subsets, targeted clustering, feature selection, mixtures of numeric and categor- ical variables, gene expression microarray data, genomics, bioinformatics 1 Introduction The goal of cluster analysis is to partition a data set of N objects into subgroups such that those in each particular group are more similar to each other than to those of other groups. DeÖning an ìencoderîfunction c(i) that maps each object i to a particular group Gl (1 l L) c(i) = l )i 2G ; (1) l one can formalize this goal as Önding the ìoptimalîencoder c (i) that minimizes a crit