Mitigating the effect of dataset shift in clustering

作者：

Highlights：

• We propose a novel kernel method for clustering in the presence of dataset shift.

• We introduce weights in the kernel K-means formulation.

• We propose a general time-dependent weighting function using OWA operators.

• The goal of the method is to prioritize recent examples over outdated ones.

• Best performance is achieved in experiments on time-dependent datasets

摘要

•We propose a novel kernel method for clustering in the presence of dataset shift.•We introduce weights in the kernel K-means formulation.•We propose a general time-dependent weighting function using OWA operators.•The goal of the method is to prioritize recent examples over outdated ones.•Best performance is achieved in experiments on time-dependent datasets

论文关键词：Induced ordered weighted average,Kernel k-means,OWA operators,Dataset shift,Clustering

论文评审过程：Received 12 November 2021, Revised 27 May 2022, Accepted 20 September 2022, Available online 23 September 2022, Version of Record 29 September 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.109058