Using Categorical Attributes for Clustering

The traditional clustering algorithms focused on clustering numeric data by exploiting the inherent geometric properties of the dataset for calculating distance functions between the points to be clustered. The distance based approach did not fit into clustering real life data containing categorical values. The focus of research then shifted to clustering such data and various categorical clustering algorithms are proposed till date. The clustering of categorical data turns complex because of the absence of a natural order on the individual domains, high dimensionality of data and the existence of subspace clusters in the categorical datasets. This survey focuses on the shortcomings of categorical data and the recent developments in the direction of using data with categorical attributes for clustering