简介:Formanyclusteringalgorithms,itisveryimportanttodetermineanappropriatenumberofclusters,whichiscalledclustervalidityproblem.Inthispaper,anewclusteringvalidityassessmentindexisproposedbasedonanovelmethodtoselectthemarginpointbetweentwoclustersforin-ter-clustersimilaritymoreaccurately,andprovidesanimprovedscatterfunctionforintra-clustersimilarity.Simulationresultsshowtheeffectivenessoftheproposedindexonthedatasetsunderconsiderationregardlessofthechoiceofaclusteringalgorithm.