Title: An Efficient Way for Hierarchical Clustering on Categorical Data Using
R Package nomclust

Authors: Zdeněk Šulc, Jaroslav Horníček, Hana Řezanková, Jana Cibulková

Abstract:
The nomclust is an R package for the hierarchical clustering of objects
characterised by nominal variables, which is available on CRAN. The package
covers the whole clustering process, i.e., calculation of a dissimilarity
matrix, application of a given hierarchical clustering method, and evaluation
of the obtained clusters. For the hierarchical clustering part, the AGNES
algorithm from the cluster package is used.

In the package, a researcher can choose from sixteen similarity measures to
determine dissimilarities between objects. The nomclust package also offers
visualisation methods of the clustering process using a dendrogram and
visualisation of evaluation criteria values to indicate the optimal number of
the final clusters. The package is well-optimised since it uses C++ code for
computationally demanding tasks, which is a benefit when the user deals with a
dataset with many variables or objects. 

The current talk will describe the possibilities of the nomclust package. The
first part will cover the nomclust package's functionalities (similarity
measures, evaluation criteria, graphical outputs, generic functions). After
this theoretical introduction, the package’s functionalities will be
demonstrated in practical examples. The final section will focus on some
thoughts about future development.