Adaptive kernel density-based anomaly detection for nonlinear systems
作者：
Highlights：
• A density-based anomaly detection approach (Adaptive-KD) to nonlinear systems is proposed.
• The approach adopts a smooth kernel function to achieve smoothness in the measure of outlierness.
• The approach applies adaptively locality-dependent kernel width to enhance its discriminating power.
• The proposed Adaptive-KD approach displays superior accuracy over other alternatives.
• Numerical illustration approves its efficacy in fault detection applications.
摘要
•A density-based anomaly detection approach (Adaptive-KD) to nonlinear systems is proposed.•The approach adopts a smooth kernel function to achieve smoothness in the measure of outlierness.•The approach applies adaptively locality-dependent kernel width to enhance its discriminating power.•The proposed Adaptive-KD approach displays superior accuracy over other alternatives.•Numerical illustration approves its efficacy in fault detection applications.
论文关键词：Unsupervised learning,Fault detection,Nonlinear systems,Kernel function,Local density
论文评审过程：Received 10 May 2017, Revised 4 October 2017, Accepted 6 October 2017, Available online 12 October 2017, Version of Record 13 November 2017.
论文官网地址：https://doi.org/10.1016/j.knosys.2017.10.009
