A Feature Selection Method based on Tree Decomposition of Correlation Graph

Abdelkader Ouali 1 Nyoman Juniarta 1 Bernard Maigret 1 Amedeo Napoli 1
1 ORPAILLEUR - Knowledge representation, reasonning
Inria Nancy - Grand Est, LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery
Abstract : This paper presents a new method for feature selection where only relevant features are kept in the dataset and all other features are discarded. The proposed method uses tree decomposition heuristics to reveal subsets of highly connected features. These subsets are replaced by selecting representatives to reduce feature redundancy. Experiments performed on various datasets show promising results for our proposals.
Complete list of metadatas

Cited literature [8 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-02194229
Contributor : Abdelkader Ouali <>
Submitted on : Thursday, July 25, 2019 - 12:34:19 PM
Last modification on : Monday, September 9, 2019 - 11:18:03 AM

File

ecml_pkdd_2019_sub11.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02194229, version 1

Collections

Citation

Abdelkader Ouali, Nyoman Juniarta, Bernard Maigret, Amedeo Napoli. A Feature Selection Method based on Tree Decomposition of Correlation Graph. LEG@ECML-PKDD 2019 - The third International Workshop on Advances in Managing and Mining Large Evolving Graphs, Sep 2019, Würzburg, Germany. ⟨hal-02194229⟩

Share

Metrics

Record views

38

Files downloads

36