Marking and Generalization by Symbolic Objects in the Symbolic Official Data Analysis Softw.pdfVIP

Marking and Generalization by Symbolic Objects in the Symbolic Official Data Analysis Software Mireille Gettler-Summa1 Lise ceremade Universit´e Paris IX Dauphine 1, Place du Ml De Lattre de Tassigny75016 Paris France Abstract. In this paper we propose an automatic method of generating Symbolic Objects in the following framework: description of a partition by symbolic objects that takes into account two aspects, that may be called homogeneity and discrimination criteria. This method belongs to a family of algorithms named MGS (Marking and Generalization by Sym- bolic Objects) in [GETSUM98], which may be applied either to Factorial Analysis interpretation in [GETSUM92][VERGIOGETSUM97], to inter- pretation of partitions [GETSUMPERFER94], or to summarizing huge databases in 1 MGS in SODAS 1.1 The input classical data matrix n observations on p nominal variables We consider a set = 1 of n objects observed for p classical nominal variables : () = is the domain of the variable , that is a finite set of categories example : = colour none, blue, red, yellow, pink n observations on p quantitative variables For the symbolic object ex- traction method which is developed in this paper, only nominal variables are accepted. Quantitative variables should be transformed into nominal ones. If no expert bounds for the intervals exist,an automatic coding is necessary. As all Marking approaches take into account a discrimination criterion (see section and section), the quality of the Markings depends on the chosen bounds for the intervals which will define the