Clustering ensembles: A hedonic game theoretical approach
作者：
Highlights：
• We investigate a new clustering ensemble algorithm (HGCE) based on hedonic games.
• HGCE can always produce a Nash and individually stable consensus partition as final solution.
• We compare HGCE with other well-known clustering ensemble methods on several data sets.
• HGCE ranked first in the statistical comparisons conducted with two external validity measures.
• HGCE is stable to perturbations to the base partitions and converged fast in all experiments.
摘要
•We investigate a new clustering ensemble algorithm (HGCE) based on hedonic games.•HGCE can always produce a Nash and individually stable consensus partition as final solution.•We compare HGCE with other well-known clustering ensemble methods on several data sets.•HGCE ranked first in the statistical comparisons conducted with two external validity measures.•HGCE is stable to perturbations to the base partitions and converged fast in all experiments.
论文关键词：Data clustering,Clustering ensemble,Hedonic game,Nash stability,Evidence accumulation
论文评审过程：Received 16 May 2017, Revised 9 March 2018, Accepted 20 March 2018, Available online 26 March 2018, Version of Record 6 April 2018.
论文官网地址：https://doi.org/10.1016/j.patcog.2018.03.017
