Deep data fuzzy clustering

摘要

In this paper we present a clustering method called Deep Data clustering. The idea of the proposed method is based on a decomposition of an input dataset. The aim od the decomposition (or dimensionality reduction) process is to reveal internal data structures in the dataset. Two methods are selected for this purpose: the principal component analysis (PCA) and the Fisher linear discriminant (FLD). The reduction process is repeated as long as the number of features is equal to one. Meanwhile, the clustering procedure is applied for the each reduced dataset. Finally, based on the clustering results obtained for the reduced datasets, the input dataset is clustered by applying the collaborative fuzzy clustering method. The well known Pima and Iris databases are used in conducted numerical experiment. The obtained results show usefulness of the proposed approach.