您好,欢迎访问三七文档
当前位置:首页 > 行业资料 > 国内外标准规范 > dm2-intro-machine-learning-classification
MachineLearning:findingpatterns2OutlineMachinelearningandClassificationExamples*LearningasSearchBiasWeka3FindingpatternsGoal:programsthatdetectpatternsandregularitiesinthedataStrongpatternsgoodpredictionsProblem1:mostpatternsarenotinterestingProblem2:patternsmaybeinexact(orspurious)Problem3:datamaybegarbledormissing4MachinelearningtechniquesAlgorithmsforacquiringstructuraldescriptionsfromexamplesStructuraldescriptionsrepresentpatternsexplicitlyCanbeusedtopredictoutcomeinnewsituationCanbeusedtounderstandandexplainhowpredictionisderived(maybeevenmoreimportant)Methodsoriginatefromartificialintelligence,statistics,andresearchondatabaseswitten&eibe5Canmachinesreallylearn?Definitionsof“learning”fromdictionary:Togetknowledgeofbystudy,experience,orbeingtaughtTobecomeawarebyinformationorfromobservationTocommittomemoryTobeinformedof,ascertain;toreceiveinstructionDifficulttomeasureTrivialforcomputersThingslearnwhentheychangetheirbehaviorinawaythatmakesthemperformbetterinthefuture.Operationaldefinition:Doesaslipperlearn?Doeslearningimplyintention?witten&eibe6ClassificationLearnamethodforpredictingtheinstanceclassfrompre-labeled(classified)instancesManyapproaches:Regression,DecisionTrees,Bayesian,NeuralNetworks,...Givenasetofpointsfromclasseswhatistheclassofnewpoint?7Classification:LinearRegressionLinearRegressionw0+w1x+w2y=0Regressioncomputeswifromdatatominimizesquarederrorto‘fit’thedataNotflexibleenough8Classification:DecisionTreesXYifX5thenblueelseifY3thenblueelseifX2thengreenelseblue5239Classification:NeuralNetsCanselectmorecomplexregionsCanbemoreaccurateAlsocanoverfitthedata–findpatternsinrandomnoise10OutlineMachinelearningandClassificationExamples*LearningasSearchBiasWeka11TheweatherproblemOutlookTemperatureHumidityWindyPlaysunnyhothighfalsenosunnyhothightruenoovercasthothighfalseyesrainymildhighfalseyesrainymildnormalfalseyesrainymildnormaltruenoovercastmildnormaltrueyessunnymildhighfalsenosunnymildnormalfalseyesrainymildnormalfalseyessunnymildnormaltrueyesovercastmildhightrueyesovercasthotnormalfalseyesrainymildhightruenoGivenpastdata,CanyoucomeupwiththerulesforPlay/NotPlay?Whatisthegame?12TheweatherproblemGiventhisdata,whataretherulesforplay/notplay?OutlookTemperatureHumidityWindyPlaySunnyHotHighFalseNoSunnyHotHighTrueNoOvercastHotHighFalseYesRainyMildNormalFalseYes……………13TheweatherproblemConditionsforplayingOutlookTemperatureHumidityWindyPlaySunnyHotHighFalseNoSunnyHotHighTrueNoOvercastHotHighFalseYesRainyMildNormalFalseYes……………Ifoutlook=sunnyandhumidity=highthenplay=noIfoutlook=rainyandwindy=truethenplay=noIfoutlook=overcastthenplay=yesIfhumidity=normalthenplay=yesIfnoneoftheabovethenplay=yeswitten&eibe14WeatherdatawithmixedattributesOutlookTemperatureHumidityWindyPlaysunny8585falsenosunny8090truenoovercast8386falseyesrainy7096falseyesrainy6880falseyesrainy6570truenoovercast6465trueyessunny7295falsenosunny6970falseyesrainy7580falseyessunny7570trueyesovercast7290trueyesovercast8175falseyesrainy7191trueno15WeatherdatawithmixedattributesHowwilltheruleschangewhensomeattributeshavenumericvalues?OutlookTemperatureHumidityWindyPlaySunny8585FalseNoSunny8090TrueNoOvercast8386FalseYesRainy7580FalseYes……………16WeatherdatawithmixedattributesRuleswithmixedattributesOutlookTemperatureHumidityWindyPlaySunny8585FalseNoSunny8090TrueNoOvercast8386FalseYesRainy7580FalseYes……………Ifoutlook=sunnyandhumidity83thenplay=noIfoutlook=rainyandwindy=truethenplay=noIfoutlook=overcastthenplay=yesIfhumidity85thenplay=yesIfnoneoftheabovethenplay=yeswitten&eibe17ThecontactlensesdataAgeSpectacleprescriptionAstigmatismTearproductionrateRecommendedlensesYoungMyopeNoReducedNoneYoungMyopeNoNormalSoftYoungMyopeYesReducedNoneYoungMyopeYesNormalHardYoungHypermetropeNoReducedNoneYoungHypermetropeNoNormalSoftYoungHypermetropeYesReducedNoneYoungHypermetropeYesNormalhardPre-presbyopicMyopeNoReducedNonePre-presbyopicMyopeNoNormalSoftPre-presbyopicMyopeYesReducedNonePre-presbyopicMyopeYesNormalHardPre-presbyopicHypermetropeNoReducedNonePre-presbyopicHypermetropeNoNormalSoftPre-presbyopicHypermetropeYesReducedNonePre-presbyopicHypermetropeYesNormalNonePresbyopicMyopeNoReducedNonePresbyopicMyopeNoNormalNonePresbyopicMyopeYesReducedNonePresbyopicMyopeYesNormalHardPresbyopicHypermetropeNoReducedNonePresbyopicHypermetropeNoNormalSoftPresbyopicHypermetropeYesReducedNonePresbyopicHypermetropeYesNormalNonewitten&eibe18AcompleteandcorrectrulesetIftearproductionrate=reducedthenrecommendation=noneIfage=youngandastigmatic=noandtearproductionrate=normalthenrecommendation=softIfage=pre-presbyopicandastigmatic=noandtearproductionrate=normalthenrecommendation=softIfage=presbyopicandspectacleprescription=myopeandastigmatic=nothenrecommendation=noneIfspectacleprescription=hypermetropeandastigmatic=noandtearproductionrate=normalthenrecommendation=softIfspectacleprescription=myopeandastigmatic=yesandtearproductionrate=normalthenrecommendation=hardIfageyoungandastigmatic=yesandtearproductionrate=normalthenrecommendation=hardIfage=pre-presby
本文标题:dm2-intro-machine-learning-classification
链接地址:https://www.777doc.com/doc-6332069 .html