A rough sets based pruning method for bagging ensemble

Ensembletechniquestrainasetofcomponentclassifiersandthencombinetheirpredictionstoclassifynewpatterns.Baggingisoneofthemostpopularensembletechniquesforimprovingweakclassifiers.However,itishardtodeployinmanyrealapplicationsbecauseofthelargememoryrequirementandhighcomputationcosttostoreandvotethepredictionsofcomponentclassifiers.Roughsettheoryisaformalmathematicaltooltodealwithincompleteorimpreciseinformation,whichhasattractedalotofattentionfromtheoryandapplicationfields.Inthispaper,anovelroughsetsbasedmethodisproposedtoprunetheclassifiersobtainedfrombaggingensembleandselectasubsetofthecomponentclassifiersforaggregation.Experimentresultsshowthattheproposedmethodnotonlydecreasesthenumberofcomponentclassifiersbutalsoobtainsacceptableperformance.