Oversampling method using outlier detectable generative adversarial network

作者：

Highlights：

• We proposed a method for balancing binary class imbalanced data using OD-GAN.

• We can detect and remove outliers with the discriminator.

• The generator generates artificial data that is very similar to the real data.

• OD-GAN showed the best performance of all the data balancing methods.

• The performance of OD-GAN increases as the outlier ratio increases.

摘要

•We proposed a method for balancing binary class imbalanced data using OD-GAN.•We can detect and remove outliers with the discriminator.•The generator generates artificial data that is very similar to the real data.•OD-GAN showed the best performance of all the data balancing methods.•The performance of OD-GAN increases as the outlier ratio increases.

论文关键词：Class imbalance problem,Oversampling,Generative adversarial network,Outlier detection

论文评审过程：Received 22 November 2018, Revised 7 May 2019, Accepted 7 May 2019, Available online 9 May 2019, Version of Record 15 May 2019.

论文官网地址：https://doi.org/10.1016/j.eswa.2019.05.006