Improved Rough Set Algorithms for Optimal Attribute Reduct.pdfVIP

1O8 JOURNALOFELECTRONICSCIENCEAND TECHNOLOG VOL．9，NO．2，JUNE2011 ImprovedRoughSetAlgorithmsforOptimal AttributeReduct c．Velayutham andK．Thangavel Abstracf-_—Featureselection(FS)aimstodeterminea evaluation function．The process aims to detemr ine a minimalfeature (attribute)subsetfrom a problem minimalfeature subsetrfom a problem domain while domain while retaining a suitably high accuracy in retaining a suitably high accuracy in representing the representing the originalfeatures．Rough settheory origina1features．Inknowledgediscovery．FSmethodsare rRST)hasbeenusedassuchatoolwithmuchsuccess． particularlydesirableasthesefacilitatetheinterpreatbility RSTenablesthediscoveryofdatadependenciesandthe oftheresultingknowledge．RoughsethteoryrRST)has reduction ofthe numberofattributescontained in a beenusedassuchatoolwithmuchsuccess．enablinghte datasetusing the data alone，requiring no additional discovery ofdatadependenciesand thereduction ofthe information．Thispaperdescribesthefundamentalideas numberoffeaturescontained in adatasetusing the daat behind RST-based approaches， reviews related FS alone．requiringnoadditionalinformation． methods built on these ideas，and analyses more TherestofthePal：Ierisorganizedasfollows．Section2 rfequently used RST-based traditionalFS algorithms presentsintroduction to therough settheory．Section 3 suchasQuickreductalgorithm，entropybasedreduct describesthe FS algorithms．Section 4 describes the algorithm，andrelativereductalgorithm．Itisfoundthat proposedmethodsforFS．The experimentalresultsare someofthedrawbacksin theexistingalgorithmsand discussed jn Section 5 nad conclusion ispresented in ourproposedimproved algorithmscanovercomethese