DBSMOTE: Density-based synthetic minority over-sampling technique

A dataset exhibits the class imbalance problem when a target class has a very small number of instances relative to other classes. A trivial classifier typically fails to detect a minority class due to its extremely low incidence rate. In this paper, a new over-sampling technique called DBSMOTE is p...

Full description

Saved in:
Bibliographic Details
Main Authors: Bunkhumpornpat,C., Sinapiromsaran,K., Lursinsap,C.
Format: Article
Published: Springer Netherlands 2015
Subjects:
Online Access:http://www.scopus.com/inward/record.url?partnerID=HzOxMe3b&scp=84862140885&origin=inward
http://cmuir.cmu.ac.th/handle/6653943832/38631
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Chiang Mai University
id th-cmuir.6653943832-38631
record_format dspace
spelling th-cmuir.6653943832-386312015-06-16T07:53:41Z DBSMOTE: Density-based synthetic minority over-sampling technique Bunkhumpornpat,C. Sinapiromsaran,K. Lursinsap,C. Artificial Intelligence A dataset exhibits the class imbalance problem when a target class has a very small number of instances relative to other classes. A trivial classifier typically fails to detect a minority class due to its extremely low incidence rate. In this paper, a new over-sampling technique called DBSMOTE is proposed. Our technique relies on a density-based notion of clusters and is designed to over-sample an arbitrarily shaped cluster discovered by DB-SCAN. DBSMOTE generates synthetic instances along a shortest path from each positive instance to a pseudo-centroid of a minority-class cluster. Consequently, these synthetic instances are dense near this centroid and are sparse far from this centroid. Our experimental results show that DBSMOTE improves precision, F-value, and AUC more effectively than SMOTE, Borderline-SMOTE, and Safe-Level-SMOTE for imbalanced datasets. © 2011 Springer-Verlag. 2015-06-16T07:53:41Z 2015-06-16T07:53:41Z 2012-04-01 Article 0924669X 2-s2.0-84862140885 10.1007/s10489-011-0287-y http://www.scopus.com/inward/record.url?partnerID=HzOxMe3b&scp=84862140885&origin=inward http://cmuir.cmu.ac.th/handle/6653943832/38631 Springer Netherlands
institution Chiang Mai University
building Chiang Mai University Library
country Thailand
collection CMU Intellectual Repository
topic Artificial Intelligence
spellingShingle Artificial Intelligence
Bunkhumpornpat,C.
Sinapiromsaran,K.
Lursinsap,C.
DBSMOTE: Density-based synthetic minority over-sampling technique
description A dataset exhibits the class imbalance problem when a target class has a very small number of instances relative to other classes. A trivial classifier typically fails to detect a minority class due to its extremely low incidence rate. In this paper, a new over-sampling technique called DBSMOTE is proposed. Our technique relies on a density-based notion of clusters and is designed to over-sample an arbitrarily shaped cluster discovered by DB-SCAN. DBSMOTE generates synthetic instances along a shortest path from each positive instance to a pseudo-centroid of a minority-class cluster. Consequently, these synthetic instances are dense near this centroid and are sparse far from this centroid. Our experimental results show that DBSMOTE improves precision, F-value, and AUC more effectively than SMOTE, Borderline-SMOTE, and Safe-Level-SMOTE for imbalanced datasets. © 2011 Springer-Verlag.
format Article
author Bunkhumpornpat,C.
Sinapiromsaran,K.
Lursinsap,C.
author_facet Bunkhumpornpat,C.
Sinapiromsaran,K.
Lursinsap,C.
author_sort Bunkhumpornpat,C.
title DBSMOTE: Density-based synthetic minority over-sampling technique
title_short DBSMOTE: Density-based synthetic minority over-sampling technique
title_full DBSMOTE: Density-based synthetic minority over-sampling technique
title_fullStr DBSMOTE: Density-based synthetic minority over-sampling technique
title_full_unstemmed DBSMOTE: Density-based synthetic minority over-sampling technique
title_sort dbsmote: density-based synthetic minority over-sampling technique
publisher Springer Netherlands
publishDate 2015
url http://www.scopus.com/inward/record.url?partnerID=HzOxMe3b&scp=84862140885&origin=inward
http://cmuir.cmu.ac.th/handle/6653943832/38631
_version_ 1681421508622155776