Using spectral entropy and bernoulli map to handle concept drift
作者：
Highlights：
• SEDD: a drift detector that considers temporal dependencies on error streams.
• SEDD detects concept drifts based on Spectral Entropy and uses the Bernoulli Map.
• SEDD is not restricted to scenarios that assume data on the stream are i.i.d.
• SEDD was tested against state of the art detectors using two base learners.
• Experiments confirmed SEDD was competitive with state-of-the-art methods.
摘要
•SEDD: a drift detector that considers temporal dependencies on error streams.•SEDD detects concept drifts based on Spectral Entropy and uses the Bernoulli Map.•SEDD is not restricted to scenarios that assume data on the stream are i.i.d.•SEDD was tested against state of the art detectors using two base learners.•Experiments confirmed SEDD was competitive with state-of-the-art methods.
论文关键词：Concept drift,Drift detection,Spectral entropy,Bernoulli map,Data stream,Online learning
论文评审过程：Received 30 March 2020, Revised 6 July 2020, Accepted 9 October 2020, Available online 23 October 2020, Version of Record 10 February 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2020.114114
