clustDDist: Clustering Discrete Distributions

Clustering of units described with distributions is considered. Frequent approach for clustering such data combines non-hierarchical method (to allow clustering of large amount of units) with hierarchical clustering method (to build dendrogram from the obtained nonhierarchical clusters and determine the most 'natural' final clustering(s) from it). The use of the squared Euclidean distance as an error function favors patterns of distributions that have one steep high peak. Here several alternative error functions are implemented. They characterize errors between clustered units and a cluster representative - leader (which needs not be defined in the same space). For these error functions the adapted leaders methods and compatible agglomerative hierarchical clustering methods are implemented.

Package details

AuthorNatasa Kejzar, Vladimir Batagelj, Simona Korenjak-Cerne
MaintainerNatasa Kejzar <natasa.kejzar@fdv.uni-lj.si>
LicenseGPL-2
Version1.3
Package repositoryView on R-Forge
Installation Install the latest version of this package by entering the following in R:
install.packages("clustDDist", repos="http://R-Forge.R-project.org")

Try the clustDDist package in your browser

Any scripts or data that you put into this service are public.

clustDDist documentation built on May 2, 2019, 6:47 p.m.