Mining for Gradualness Over Time Using Sequential Patterns

Lisa Di Jorio 1 Anne Laurent 1 Maguelonne Teisseire 1
1 TATOO - Fouille de données environnementales
LIRMM - Laboratoire d'Informatique de Robotique et de Microélectronique de Montpellier
Abstract : Sequential patterns have been studied for the last fifteen years and are now well-recognized as a key method for extracting relevant knowledge from large databases. They have been extended to handle quantitative attributes, both in the crisp and fuzzy context. However, they still fail to catch one of the key information often hidden in the data, i.e. graduality. In this paper, we identify two kinds of gradualness that can be combined with the ordering notion of sequential patterns. For instance, gradualness can be seen as a {\bf temporal evolution} or as a {\bf values evolution} from one object to another one. We then propose an efficient algorithm to handle the first kind of gradual sequential pattern. We also provide the necessary theoretical material.
Document type :
Conference papers
Complete list of metadatas

https://hal-lirmm.ccsd.cnrs.fr/lirmm-00362568
Contributor : Lisa Di Jorio <>
Submitted on : Wednesday, February 18, 2009 - 4:41:10 PM
Last modification on : Friday, October 19, 2018 - 1:14:12 AM

Identifiers

  • HAL Id : lirmm-00362568, version 1

Collections

Citation

Lisa Di Jorio, Anne Laurent, Maguelonne Teisseire. Mining for Gradualness Over Time Using Sequential Patterns. KES-IDT'09: First KES International Symposium on Intelligent Decision Technologies, Apr 2009, Himeji, Japan, pp.000-010. ⟨lirmm-00362568⟩

Share

Metrics

Record views

100