1. Introduction
Reducing multi-class problems by binarization techniques
3. State of the art on aggregation schemes for binarization
techniques
Experimental framework.
Experimental study.
In thiswork,we are interested in ensemble methods by binarization techniques; in particular,we
focus on the well-known one-vs-one and one-vs-all decomposition strategies,paying special attention
to the fina lstep of the ensembles,the combination of the outputs of the binary classifiers.
Ouraimisto
developanempiricalanalysisofdifferentaggregationstocombinetheseoutputs.Todoso,wedevelop
a doublestudy:first,weusedifferentbaseclassifiersinordertoobservethesuitabilityandpotentialof
each combinationwithineachclassifier.Then,wecomparetheperformanceoftheseensemble
techniqueswiththeclassifiers’themselves.Hence,wealsoanalysetheimprovementwithrespectto
the classifiersthathandlemultipleclassesinherently.
We carryouttheexperimentalstudywithseveralwell-knownalgorithmsoftheliteraturesuchas
Support VectorMachines,DecisionTrees,InstanceBasedLearningorRuleBasedSystems.Wewillshow,
supportedbyseveralstatisticalanalyses,thegoodnessofthebinarizationtechniqueswithrespecttothe
baseclassifiersandfinallywewillpointoutthemostrobusttechniqueswithinthisframework.
baseclassifiersandfinallywewillpointoutthemostrobusttechniqueswithinthisframework.