A COMBINED K-MEANSAND HIERARCHICAL CLUSTERINGMETHOD FOR IMPROVING THE CLUSTERINGEFFICIENCYOF MICROARRAY
Tung-Shou Chen, Tzu‐Hsin Tsai, Yi‐Tzu Chen, Chin-Chiang Lin, Rong-Chang Chen · 2005
Amongthemicroarray dataanalysis clustering methods, K-meansandhierarchical clustering areresearchers' favorable tools today. However, eachofthese traditional clustering methods hasitslimitations. Inthis study, we introduce anewmethod, hierarchical K-means regulating divisive oragglomerative approach. Thehierarchical Kmeansfirstly employs K-means' algorithm ineachcluster todetermine K cluster while operating andthenemploys itonhierarchical clustering technique toshorten merging clusters timewhile generating atree-like dendrogram. We apply this method intwooriginal microarray datasets. The result indicates divisive hierarchical K-meansissuperior tohierarchical clustering oncluster quality andissuperior toK-meansclustering on computational speed. Our conclusion is thatdivisive hierarchical K-means establishes a better clustering algorithm satisfying researchers' demand.