Robust clustering by identifying the veins of clusters based on kernel density estimation
作者：
Highlights：
• A robust clustering algorithm(IVDPC) is proposed to solve the ”chain reaction“ and cut off distance selecting problems of DPC.
• A new similarity coefficient is introduced to represent the relevance between the points which is an extension of γ defined in DPC.
• The local density is estimated through a non-parametric density estimation method so as to eliminate the reliance of user-defined parameter dc.
• Clusters are characterized by veins rather than one representative point, which allows IVDPC to identify the main structure of clusters more visualized and precise.
• The robustness of the algorithm with respect to the choice of input parameters is proved via statistical method.
摘要
•A robust clustering algorithm(IVDPC) is proposed to solve the ”chain reaction“ and cut off distance selecting problems of DPC.•A new similarity coefficient is introduced to represent the relevance between the points which is an extension of γ defined in DPC.•The local density is estimated through a non-parametric density estimation method so as to eliminate the reliance of user-defined parameter dc.•Clusters are characterized by veins rather than one representative point, which allows IVDPC to identify the main structure of clusters more visualized and precise.•The robustness of the algorithm with respect to the choice of input parameters is proved via statistical method.
论文关键词：Robust clustering,Veins of clusters,Density peaks,Kernel density estimation
论文评审过程：Received 3 July 2017, Revised 26 June 2018, Accepted 30 June 2018, Available online 7 July 2018, Version of Record 10 September 2018.
论文官网地址：https://doi.org/10.1016/j.knosys.2018.06.021
