Methods for Clustering Categorical and Mixed Data: An Overview and New Algorithms

Abstract

Methods of clustering for categorical and mixed data are considered. Dissimilarities for this purpose are reviewed and different classes of algorithms according to different classes of similarities are discussed. Details of several algorithms are then given, which include agglomerative hierarchical clustering, K-means and related methods such as K-medoids and K-modes, and methods of network clustering. The way how the combinations of existing ideas leads to new algorithms is discussed.