Skip to Main content Skip to Navigation
Theses

Extraction de connaissances dans les bases de donn'ees comportant des valeurs manquantes ou un grand nombre d'attributs

Abstract : Knowledge Discovery in Databases is a recent field aiming at discovering new knowledge. Pattern mining is here a central task and this thesis tackles two generic cases: databases containing missing values or a large number of attributes. Firstly, we propose a temporary desactivation process of the incomplete objects, which allows to lead computations in an incomplete database and gives rise to properties compatible with the complete database. An original method for building informative and generalised association rules combines the properties of the opposite database. Secondly, we have developed a complete theoretical framework for the constrained mining of patterns using a transposition principle and the \Galois connection properties. It enables to choose the most favourable orientation of the database. Search constraints are also transposable, and allow to get the constrained patterns by leading extractions in the transposed context. At the end, the use of generalised association rules for supervised learning and strong emerging patterns complete these works in both medical and genomic area.
Document type :
Theses
Complete list of metadata

https://tel.archives-ouvertes.fr/tel-00252089
Contributor : Hal System <>
Submitted on : Tuesday, February 12, 2008 - 3:56:23 PM
Last modification on : Tuesday, February 5, 2019 - 12:12:10 PM
Long-term archiving on: : Monday, May 17, 2010 - 6:37:56 PM

Identifiers

  • HAL Id : tel-00252089, version 1

Citation

François Rioult. Extraction de connaissances dans les bases de donn'ees comportant des valeurs manquantes ou un grand nombre d'attributs. Autre [cs.OH]. Université de Caen, 2005. Français. ⟨tel-00252089⟩

Share

Metrics

Record views

283

Files downloads

1987