Selection-based resampling ensemble algorithm for nonstationary imbalanced stream data learning
作者：
Highlights：
• A classifier is proposed to handle concept drift and class imbalance.
• The selectively resampling method avoids drifting examples and difficult examples.
• SRE can quickly react to different kinds of concept drift.
• Costly misclassification examples and minority examples are emphasized.
• SRE is robust against chunk size.
• SRE obtains a good overall balance in accuracy, recall, G-mean, F-measure, and AUC.
摘要
•A classifier is proposed to handle concept drift and class imbalance.•The selectively resampling method avoids drifting examples and difficult examples.•SRE can quickly react to different kinds of concept drift.•Costly misclassification examples and minority examples are emphasized.•SRE is robust against chunk size.•SRE obtains a good overall balance in accuracy, recall, G-mean, F-measure, and AUC.
论文关键词：Data stream classification,Concept drift,Class imbalance,Ensemble
论文评审过程：Received 12 May 2018, Revised 9 August 2018, Accepted 20 September 2018, Available online 3 October 2018, Version of Record 21 November 2018.
论文官网地址：https://doi.org/10.1016/j.knosys.2018.09.032
