Skip to Main content Skip to Navigation

Un cadre générique de découverte de motifs sous contraintes fondées sur des primitives

Abstract : Pattern mining is a significant field of Knowledge Discovery in
Databases. This thesis deals with the mining problem of local patterns
under constraints. We propose a new framework relying on monotone
primitives in order to define and mine varied constraints. This broad
spectrum of constraints enables users to accurately focus on the most
interesting patterns. We provide two main approaches for automatically
mining patterns which solve the intrinsic algorithmic difficulty of
this task. Their efficiency mainly relies on necessary conditions
approximating the variation of contraints. Firstly, relaxing methods
enable us to re-use numerous usual algorithms. Secondly, we design
pattern mining algorithms dedicated to wide or correlated datasets by
exploiting the concept of equivalence classes. Finally, the use of
these methods highlights several relevant local phenomena in real
industrial and medical applications.
Document type :
Complete list of metadata

Cited literature [121 references]  Display  Hide  Download
Contributor : Hal System <>
Submitted on : Monday, January 8, 2007 - 4:00:54 PM
Last modification on : Tuesday, February 5, 2019 - 12:12:10 PM
Long-term archiving on: : Tuesday, April 6, 2010 - 9:47:41 PM


  • HAL Id : tel-00123185, version 1


Arnaud Soulet. Un cadre générique de découverte de motifs sous contraintes fondées sur des primitives. Autre [cs.OH]. Université de Caen, 2006. Français. ⟨tel-00123185⟩



Record views


Files downloads