SMOTE-least square support vector machine for classification of multiclass imbalanced data

Santi Wulan Purnami, Rani Kemala Trapsilasiwi

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

10 Citations (Scopus)

Abstract

Dealing with multiclass classification problem is still considered as significant hurdle to determine an efficient classifier. Moreover, this task is getting rough when it comes to imbalanced data, which defined as the number of some classes are much bigger than the others. This condition could cause the classifier tends to predict the majority class and ignore the minority class. This study proposed Synthetic Minority Oversampling Technique-Least Square Support Vector Machine (SMOTE-LSSVM) to build a classifier addressing this problem. Particle Swarm Optimization-Gravitational Search Algorithm (PSO-GSA) was used to optimize the parameters of LS-SVM, while SMOTE was employed to balance the data. The effectiveness of SMOTE-LSSVM was examined on malignancy of breast cancer dataset. Results of this studies showed that the accuracy rate after applying SMOTE increased significantly compare to the results without applying SMOTE.

Original languageEnglish
Title of host publicationProceedings of 2017 9th International Conference on Machine Learning and Computing, ICMLC 2017
PublisherAssociation for Computing Machinery
Pages107-111
Number of pages5
ISBN (Electronic)9781450348171
DOIs
Publication statusPublished - 24 Feb 2017
Event9th International Conference on Machine Learning and Computing, ICMLC 2017 - Singapore, Singapore
Duration: 24 Feb 201726 Feb 2017

Publication series

NameACM International Conference Proceeding Series
VolumePart F128357

Conference

Conference9th International Conference on Machine Learning and Computing, ICMLC 2017
Country/TerritorySingapore
CitySingapore
Period24/02/1726/02/17

Keywords

  • Imbalanced data
  • Least square support vector machine
  • Multiclass
  • Particle swarm optimization-gravitational search algorithm
  • Synthetic minority oversampling technique

Fingerprint

Dive into the research topics of 'SMOTE-least square support vector machine for classification of multiclass imbalanced data'. Together they form a unique fingerprint.

Cite this