Adapting dynamic classifier selection for concept drift
作者：
Highlights：
• We show that most DCS methods can be adapted to concept drift scenarios.
• A time dependency is modeled according to the concept drift nature (real/virtual).
• We discuss the impact of the pool pruning and introduce the concept diversity idea.
• The DCS is tested under real and virtual concept drift scenarios.
• The PKLot dataset is used as a real world concept drift benchmark.
摘要
•We show that most DCS methods can be adapted to concept drift scenarios.•A time dependency is modeled according to the concept drift nature (real/virtual).•We discuss the impact of the pool pruning and introduce the concept diversity idea.•The DCS is tested under real and virtual concept drift scenarios.•The PKLot dataset is used as a real world concept drift benchmark.
论文关键词：Concept drift,Dynamic classifier selection,Dynamic ensemble selection,Concept diversity
论文评审过程：Received 20 December 2017, Revised 12 March 2018, Accepted 13 March 2018, Available online 15 March 2018, Version of Record 21 March 2018.
论文官网地址：https://doi.org/10.1016/j.eswa.2018.03.021
