Use of Pyramids in Symbolic Data Analysis

A method of symbolic pyramidal clustering is presented, which allows to construct a pyramid on a set of multi-valued data. The criteria guiding cluster formation are based on the notions of extension/intension, and lead to clusters allowing for a conjunctive description. Both input data and the obtained clusters are represented within the framework of the formalism of symbolic objects. The method is positionned in the context of conceptual clustering methods. Its advantages and drawbacks are then discussed on the basis of some applications.