Anytime Subgroup Discovery in Numerical Domains with Guarantees - Technical Report

Aimene Belfodil 1 Adnene Belfodil 2, 1 Mehdi Kaytoue 1, 3
1 DM2L - Data Mining and Machine Learning
LIRIS - Laboratoire d'InfoRmatique en Image et Systèmes d'information
2 BD - Base de Données
LIRIS - Laboratoire d'InfoRmatique en Image et Systèmes d'information
Abstract : Subgroup discovery is the task of discovering patterns that accurately discriminate a class label from the others. Existing approaches can uncover such patterns either through an exhaustive or an approximate exploration of the pattern search space. However, an exhaustive exploration is generally unfeasible whereas approximate approaches do not provide guarantees bounding the error of the best pattern quality nor the exploration progression ("How far are we of an exhaustive search"). We design here an algorithm for mining numerical data with three key properties w.r.t. the state of the art: (i) It yields progressively interval patterns whose quality improves over time; (ii) It can be interrupted anytime and always gives a guarantee bounding the error on the top pattern quality and (iii) It always bounds a distance to the exhaustive exploration. After reporting experimentations showing the effectiveness of our method, we discuss its generalization to other kinds of patterns.
Document type :
Reports
Complete list of metadatas

Cited literature [31 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01874949
Contributor : Adnene Belfodil <>
Submitted on : Saturday, September 15, 2018 - 11:43:55 AM
Last modification on : Thursday, February 14, 2019 - 1:14:13 AM
Long-term archiving on : Sunday, December 16, 2018 - 12:29:36 PM

File

Anytime Subgroup Discovery in ...
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01874949, version 1

Citation

Aimene Belfodil, Adnene Belfodil, Mehdi Kaytoue. Anytime Subgroup Discovery in Numerical Domains with Guarantees - Technical Report. [Technical Report] LIRIS UMR 5205 CNRS/INSA de Lyon/Université Claude Bernard Lyon 1/Université Lumière Lyon 2/École Centrale de Lyon. 2019. ⟨hal-01874949⟩

Share

Metrics

Record views

360

Files downloads

143