Clustering of Symbolic Data based on Affinity Coefficient: Application to real data sets

Abstract

The increasing use of databases, often large ones, in diverse areas of study makes it pertinent to summarise data in terms of their most relevant concepts. These concepts may be described by types of complex data, also known as symbolic data […]. We present some results from the Ascendant Hierarchical Cluster Analysis (AHCA) of symbolic objects described by interval data, in order to illustrate the effectiveness of the Ascendent Hierarchical Cluster Analysis based on the weighted generalized affinity coefficient, for symbolic data. The measure of comparison between the elements was combined with classical aggregation criteria and probabilistic ones. The probabilistic aggregation criteria used in this study belong to a parametric family of methods in the scope of the probabilistic approach of AHCA, named VL methodology and the validation of the clustering results is based on some validation measures. Finally, we compare the results achieved by our approach with the ones obtained by other authors.