Skip to Main content Skip to Navigation

Classification automatique et modèles

Abstract : Up to now, two parallel trends have emerged in the developement and practice of statistical data processing. The first one involves methods that consider the possibility of a probalilistic interpretation ; the second one uses a rather large group of automatic clustering methods applied within a purely geometrical framework. Our study is set halfway betwen those two approach ; indeed the links that exist betwen the probabilistic approach and the geometrical approach have enabled us to interpret automatic clustering methods in probabilistic terms, to propose new criteria that can improve the quality of the partition ; we then extend the study of these links to cases were the data involve two sets ; we show how the cross clustering can be seen as a solution to a problem for the estimation of the parameters of a model with crossed mixture, we develop a method of identification of crossed mixture ; this method will enable us to interpret cross clustering methods and to propose new cross clustering algorithms using adaptative distances. Some methods proposed in this study have been programmed and integrated into the data analysis software SICLA (Interactif Systeme of Automatic Clustering, INRIA)
Document type :
Complete list of metadata

Cited literature [19 references]  Display  Hide  Download
Contributor : Administrateur Du Ccsd <>
Submitted on : Tuesday, April 24, 2018 - 3:48:07 PM
Last modification on : Thursday, April 26, 2018 - 1:28:28 AM
Long-term archiving on: : Wednesday, September 19, 2018 - 12:07:00 PM


Files produced by the author(s)


  • HAL Id : tel-01775952, version 1



Yamina Bencheikh. Classification automatique et modèles. Mathématiques générales [math.GM]. Université Paul Verlaine - Metz, 1992. Français. ⟨NNT : 1992METZ002S⟩. ⟨tel-01775952⟩



Record views


Files downloads