CDF Transform-and-Shift: An effective way to deal with datasets of inhomogeneous cluster densities

作者：

Highlights：

• Addressing the long-standing inhomogeneous cluster density issue.

• Analysing the drawbacks of existing approaches.

• Proposing a new method to homogenise cluster density while preserving cluster structure.

• Demonstrate its effectiveness in existing clustering algorithms and anomaly detectors.

摘要

•Addressing the long-standing inhomogeneous cluster density issue.•Analysing the drawbacks of existing approaches.•Proposing a new method to homogenise cluster density while preserving cluster structure.•Demonstrate its effectiveness in existing clustering algorithms and anomaly detectors.

论文关键词：Density-ratio,Density-based clustering,kNN Anomaly detection,Inhomogeneous cluster densities,Scaling,Shift

论文评审过程：Received 7 September 2020, Revised 15 February 2021, Accepted 31 March 2021, Available online 8 April 2021, Version of Record 18 April 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107977