Algorithms for Internal Validation Clustering Measures in the Post Genomic Era

摘要: Inferring cluster structure in microarray datasets is a fundamental task for the -omic sciences. A question Statistics, Data Analysis and Classification, prediction of number clusters dataset, usually established via internal validation measures. Despite wealth measures available literature, new ones have been recently proposed, some them specifically data. In this dissertation, study given, paying particular attention to stability based ones. Indeed, class particularly prominent promising order reliable estimate dataset. For those measures, general algorithmic paradigm proposed here that highlights richness accounts already literature. Moreover, most representative are also considered. Experiments on 12 benchmark performed assess both intrinsic ability measure predict correct dataset its merit relative other The main result hierarchy terms precision speed, highlighting their merits limitations not reported before This shows faster measure, less accurate it is. reduce time performance gap between fastest precise technique designing fast approximation algorithms systematically applied. end speed-up many studied brings within one magnitude time, with no degradation power. Prior work, was at least two orders magnitude.