Geometric SMOTE a geometrically enhanced drop-in replacement for SMOTE

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Classification of imbalanced datasets is a challenging task for standard algorithms. Although many methods exist to address this problem in different ways, generating artificial data for the minority class is a more general approach compared to algorithmic modifications. SMOTE algorithm, as well as any other oversampling method based on the SMOTE mechanism, generates synthetic samples along line segments that join minority class instances. In this paper we propose Geometric SMOTE (G-SMOTE) as a enhancement of the SMOTE data generation mechanism. G-SMOTE generates synthetic samples in a geometric region of the input space, around each selected minority instance. While in the basic configuration this region is a hyper-sphere, G-SMOTE allows its deformation to a hyper-spheroid. The performance of G-SMOTE is compared against SMOTE as well as baseline methods. We present empirical results that show a significant improvement in the quality of the generated data when G-SMOTE is used as an oversampling algorithm. An implementation of G-SMOTE is made available in the Python programming language.

Original languageEnglish
Pages (from-to)118-135
Number of pages18
JournalInformation Sciences
Volume501
DOIs
Publication statusPublished - 1 Oct 2019

Fingerprint

Replacement
Oversampling
Computer programming languages
Python
Line segment
Minorities
Join
Programming Languages
Baseline
Enhancement
Configuration
Empirical results
Programming
Language

Keywords

  • Classification
  • Data generation
  • Imbalanced learning
  • Oversampling
  • SMOTE
  • Supervised learning

Cite this

@article{443397f800a94476af632f6b22abc845,
title = "Geometric SMOTE a geometrically enhanced drop-in replacement for SMOTE",
abstract = "Classification of imbalanced datasets is a challenging task for standard algorithms. Although many methods exist to address this problem in different ways, generating artificial data for the minority class is a more general approach compared to algorithmic modifications. SMOTE algorithm, as well as any other oversampling method based on the SMOTE mechanism, generates synthetic samples along line segments that join minority class instances. In this paper we propose Geometric SMOTE (G-SMOTE) as a enhancement of the SMOTE data generation mechanism. G-SMOTE generates synthetic samples in a geometric region of the input space, around each selected minority instance. While in the basic configuration this region is a hyper-sphere, G-SMOTE allows its deformation to a hyper-spheroid. The performance of G-SMOTE is compared against SMOTE as well as baseline methods. We present empirical results that show a significant improvement in the quality of the generated data when G-SMOTE is used as an oversampling algorithm. An implementation of G-SMOTE is made available in the Python programming language.",
keywords = "Classification, Data generation, Imbalanced learning, Oversampling, SMOTE, Supervised learning",
author = "Georgios Douzas and Fernando Bacao",
note = "Douzas, G., & Bacao, F. (2019). Geometric SMOTE a geometrically enhanced drop-in replacement for SMOTE. Information Sciences, 501, 118-135. https://doi.org/10.1016/j.ins.2019.06.007",
year = "2019",
month = "10",
day = "1",
doi = "10.1016/j.ins.2019.06.007",
language = "English",
volume = "501",
pages = "118--135",
journal = "Information Sciences",
issn = "0020-0255",
publisher = "Elsevier Science B.V., Amsterdam.",

}

Geometric SMOTE a geometrically enhanced drop-in replacement for SMOTE. / Douzas, Georgios; Bacao, Fernando.

In: Information Sciences, Vol. 501, 01.10.2019, p. 118-135.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Geometric SMOTE a geometrically enhanced drop-in replacement for SMOTE

AU - Douzas, Georgios

AU - Bacao, Fernando

N1 - Douzas, G., & Bacao, F. (2019). Geometric SMOTE a geometrically enhanced drop-in replacement for SMOTE. Information Sciences, 501, 118-135. https://doi.org/10.1016/j.ins.2019.06.007

PY - 2019/10/1

Y1 - 2019/10/1

N2 - Classification of imbalanced datasets is a challenging task for standard algorithms. Although many methods exist to address this problem in different ways, generating artificial data for the minority class is a more general approach compared to algorithmic modifications. SMOTE algorithm, as well as any other oversampling method based on the SMOTE mechanism, generates synthetic samples along line segments that join minority class instances. In this paper we propose Geometric SMOTE (G-SMOTE) as a enhancement of the SMOTE data generation mechanism. G-SMOTE generates synthetic samples in a geometric region of the input space, around each selected minority instance. While in the basic configuration this region is a hyper-sphere, G-SMOTE allows its deformation to a hyper-spheroid. The performance of G-SMOTE is compared against SMOTE as well as baseline methods. We present empirical results that show a significant improvement in the quality of the generated data when G-SMOTE is used as an oversampling algorithm. An implementation of G-SMOTE is made available in the Python programming language.

AB - Classification of imbalanced datasets is a challenging task for standard algorithms. Although many methods exist to address this problem in different ways, generating artificial data for the minority class is a more general approach compared to algorithmic modifications. SMOTE algorithm, as well as any other oversampling method based on the SMOTE mechanism, generates synthetic samples along line segments that join minority class instances. In this paper we propose Geometric SMOTE (G-SMOTE) as a enhancement of the SMOTE data generation mechanism. G-SMOTE generates synthetic samples in a geometric region of the input space, around each selected minority instance. While in the basic configuration this region is a hyper-sphere, G-SMOTE allows its deformation to a hyper-spheroid. The performance of G-SMOTE is compared against SMOTE as well as baseline methods. We present empirical results that show a significant improvement in the quality of the generated data when G-SMOTE is used as an oversampling algorithm. An implementation of G-SMOTE is made available in the Python programming language.

KW - Classification

KW - Data generation

KW - Imbalanced learning

KW - Oversampling

KW - SMOTE

KW - Supervised learning

UR - http://www.scopus.com/inward/record.url?scp=85066855661&partnerID=8YFLogxK

UR - http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcAuth=Alerting&SrcApp=Alerting&DestApp=WOS_CPL&DestLinkType=FullRecord&UT=WOS:000480663900008

U2 - 10.1016/j.ins.2019.06.007

DO - 10.1016/j.ins.2019.06.007

M3 - Article

VL - 501

SP - 118

EP - 135

JO - Information Sciences

JF - Information Sciences

SN - 0020-0255

ER -