Cost-sensitive ensemble of stacked denoising autoencoders for class imbalance problems in business domain

作者：

Highlights：

• Novel methods that uses deep neural networks, cost-sensitive and ensemble learning.

• The methods are compared with 12 methods on 6 large real-life imbalance data sets.

• They outperforms existing methods in generalization performance.

• They obtain low generalization gaps and can avoid overfitting.

• CSDE gives excellent results on data sets across a wide range of imbalance ratios.

摘要

•Novel methods that uses deep neural networks, cost-sensitive and ensemble learning.•The methods are compared with 12 methods on 6 large real-life imbalance data sets.•They outperforms existing methods in generalization performance.•They obtain low generalization gaps and can avoid overfitting.•CSDE gives excellent results on data sets across a wide range of imbalance ratios.

论文关键词：Cost-sensitive,Stacked denoising autoencoders,Ensemble,Class imbalance

论文评审过程：Received 23 November 2018, Revised 1 September 2019, Accepted 2 September 2019, Available online 2 September 2019, Version of Record 18 September 2019.

论文官网地址：https://doi.org/10.1016/j.eswa.2019.112918