傳統(tǒng)電腦決策課件

TheApplicationofFuzzySetsonDataMiningProf.Tzung-PeiHongDepartmentofElectricalEngineeringNationalUniversityofKaohsiungTheApplicationofFuzzySetsOutlineIntroductionReviewDataMiningFuzzySetsFuzzyDataMiningFuzzyAssociationRules(I)FuzzyAssociationRules(II)FuzzyGeneralizedAssociationRulesFuzzyWebMiningOutlineIntroductionOutlineReviewofGAFuzzydataminingformembershipfunctionsandrulesTwoapproachesConclusionOutlineReviewofGAReasonsfordataminingSamGoodsSupermarketHowtoarrangegoodsintosupermarket?

ReasonsfordataminingSamGoodReasonsfordataminingSamcustomerhowtomakethemarketingstrategiesforthesecustomer?

ReasonsfordataminingSamcustMiningassociationrules

BreadMilkIFbreadisbought thenmilkisboughtMiningassociationrules

BreadTheroleofdataminingUsefulpatternsTransactiondataPreprocessdataDataMiningKnowledgeandstrategyDatainformationTheroleofdataminingUsefulDifferentkindsofknowledgeAssociationrulesGeneralizedassociationrulesSequentialpatternsQuantitativeassociationrulesClassificationrulesClusteringrulesetc…FocusDifferentkindsofknowledgeAsMiningassociationrules

BreadMilkIFbreadisbought thenmilkisboughtMiningassociationrules

BreaAprioriAlgorithm

ProposedbyAgrawaletal.Step1:Defineminsupandminconfex:minsup=50% minconf=50%Step2:FindlargeitemsetsStep3:GenerateassociationrulesAprioriAlgorithm

ProposedbyExampleLargeitemsetsScanDatabaseScanDatabaseScanDatabaseItemsetSup.{A}2{B}3{C}3{E}3L1ExampleLargeitemsetsScanScanSExample

ExampleFuzzySets傳統(tǒng)電腦決策不是對(duì)(1)就是錯(cuò)(0) 例如：25歲以上是青年，那26歲就是中年? 60分以上是及格，那60分以下就是不及格何謂模糊在對(duì)(1)與錯(cuò)(0)之間，再多加幾個(gè)等級(jí)幾乎對(duì)(0.8)可能對(duì)(0.6)可能錯(cuò)(0.4)幾乎錯(cuò)(0.2)FuzzySets傳統(tǒng)電腦決策FuzzySetsQuestion:168公分到底算不算高?身高(Cm)中矮高170180160隸屬度再多分成幾級(jí)連續(xù)FuzzySetsQuestion:168公分到底算不算高Example:“Closeto0”e.g.μA(3)=0.01μA(1)=0.09μA(0.25)=0.62μA(0)=1DefineaMembershipFunction:μA(x)=

Example:“Closeto0”e.g.Example:“Closeto0”VeryCloseto0:μA(x)=

Example:“Closeto0”VeryCloseFuzzySet(Cont.)Membershipfunction[0,1]e.g.sunny:x→[0,1]0.6sunny0.8sunny0.1sunnyxFuzzySet(Cont.)MembershipfuFuzzySetSimpleIntuitivelypleasingAgeneralizationofcrispsetVaguemember→non-memberSunnyNotsunny0.200or1Non-membermembergradualFuzzySetSunnyNotsunny1FuzzyOperations交集(AND)取較小的可能性 EX:學(xué)生聰明(0.8)而且用功(0.6)則是模範(fàn)生(0.6)聯(lián)集(OR)取較大的可能性 EX:學(xué)生聰明(0.8)或者用功(0.6)則是模範(fàn)生(0.8)反面(NOT)取與1的差 EX:學(xué)生聰明是0.8,則學(xué)生不聰明0.2FuzzyOperationsFuzzyInferenceExample洪老師找小老婆的條件(大眼睛而且小嘴巴)或者是身材好Question:誰(shuí)是最佳女主角

大眼睛 小嘴巴 身材好陶晶瑩 0 0.8 0.3張惠妹 1 0.6 0.8李玟 0 0.3 0.9李心潔 0.7 0.1 0.5蔡依林 0.8 0.5 0.3FuzzyInferenceExample洪老師找小老婆Answer對(duì)陶晶瑩=(0AND0.8)OR0.3=0OR0.3=0.3對(duì)張惠妹=(1AND0.6)OR0.8=0.8對(duì)李玟=(0AND0.3)OR0.9=0.9對(duì)李心潔=(0.7AND0.1)OR0.5=0.5對(duì)蔡依林=(0.8AND0.5)OR0.3=0.5李玟為最佳選擇!謝謝!Answer對(duì)陶晶瑩=(0AND0.8)OR0.3FuzzyDecisionA={A1,A2,A3,A4,A5}AsetofalternativesC={C1,C2,C3}AsetofcriteriaC1(bigeyes)C2(smallmouth)C3(goodshape)A1(Mary)00.80.3A2(Judy)10.60.8A3(Jan)00.30.9A4(Mandy)A5(Nancy)FuzzyDecisionA={A1,A2,A3,Example(Cont.)Assume:C1andC2orC3E(Ai):evaluationfunctionE(A1)=(00.8)0.3=00.3=0.3E(A2)=(10.6)0.8=0.60.8=0.8E(A3)=(00.3)0.9=00.9=0.9thebestchoiceE(A4)=(0.70.1)0.5=0.10.5=0.5E(A5)=(0.80.5)0.3=0.50.3=0.5C1(bigeyes)C2(smallmouth)C3(goodshape)A1(Mary)00.80.3A2(Judy)10.60.8A3(Jan)00.30.9A4(Mandy)A5(Nancy)Example(Cont.)Assume:C1andMotivationforFuzzyMiningInreal-worldapplications

Transactionswith

quantitativevaluesUsingfuzzysetstoprocessitTIDPurchaseditems1(A,3)(C,4)(E,2)2(B,3)(C,7)(D,7)3(B,2)(C,10)(E,5)4(A,9)(E,10)5(A,7)(D,8)6(B,2)(C,8)(D,10)MotivationforFuzzyMiningInFuzzyDataMiningSolvingquantitativevalues

e.g.Johnbuys10bread,2butterand3Milk.

FuzzyDataMiningMuchBreadandLittleButterMiddleAmountofMilkFuzzyDataMiningSolvingquantFuzzydataminingQuantitativedataLinguistictermMembershipfunction

01611LowMiddleHigh10MembershipvalueNumberofitemFuzzydataminingQuantitativeMainIdeaNumericDatabaseFuzzyDataMiningKnowledge61218LowMiddleHighQuantity0NumericDatabase

milkbreadcookiesbeverageT14222T27307T38539T491513If

milk.MiddleThen

cookies.LowMainIdeaNumericFuzzyDataMinRelatedresearchLeeandHyung,1997α-cutConvertingthemembershiptuplestobinarytuplesPedrycz,1996,1998

RunningtheFCM(FuzzyC-Means)methodSolvingclusterproblemRelatedresearchLeeandHyung,RelatedresearchChanandhisco-workers,1997Maddourietal.,1998Rubin,1998Hanetal.,1998UsingMachinelearningmethodCombiningdataminingRelatedresearchChanandhiscFuzzyMiningAlgorithmInputnquantitativetransactiondatamattributesAsetofmembershipfunctionsTwothresholdsMinimumsupport=Minimumconfidence=OutputAsetoffuzzyassociationrulesFuzzyMiningAlgorithmInputFuzzyMiningAlgorithmStep1Transformthequantitativevalueofeachtransactiondatumintoafuzzysetusingthegivenmembershipfunctions.Step2CalculatethescalarcardinalityofeachattributeregioninthetransactiondataFuzzyMiningAlgorithmStep1FuzzyMiningAlgorithmStep3Foreachfuzzyregion,checkwhetherit

isinthesetoflarge1-itemsets(L1)Step4Setr=1,whererisusedtorepresentthenumberofitemskeptinthecurrentlargeitemsetsFuzzyMiningAlgorithmStep3FuzzyMiningAlgorithmStep5GeneratethecandidatesetCr+1fromLrinawaysimilartothatintheapriorialgorithmexceptthattworegionsbelongingtothesameattributecannotsimultaneouslyexistinanitemsetinCr+1.FuzzyMiningAlgorithmStep5FuzzyMiningAlgorithmStep6Dothefollowingsubstepsforeachnewlyformedcandidate(r+1)-itemsetswithitemsinCr+1:Step6.1:Calculatethefuzzyvalueofeachtransactiondatains;Step6.2:Calculatethescalarcardinalityofsonthetransactions;Step6.3:Ifcounts

islargerthanorequaltothepredefinedminimumsupportvalue,putsinLr+1

FuzzyMiningAlgorithmStep6FuzzyMiningAlgorithmStep7IFLr+1isnull,thendothenextstep;otherwise,setr=r+1andrepeatSTEPs5to7.Step8Constructtheassociationrulesforalllargeq-itemsetswithitems,usingthefollowingsubsteps:Step8.1:Formallpossibleassociationrule:Step8.2:Calculatetheconfidencevaluesofallassociationrules:Step9Outputtheruleswithconfidencevalueslargerthanorequaltothepredefinedconfidencethreshold

FuzzyMiningAlgorithmStep7AnExampleGeneratingassociationrulesforcoursegradesAccordingtohistoricaldataconcerningstudents’coursescores.Thedataset10transactions.EachcaseconsistsoffivecoursescoresObject-OrientedProgramming(denotedOOP),Database(denotedDB),Statistics(denotedST),DataStructure(denotedDS),ManagementInformationSystem(denotedMIS).AnExampleGeneratingassociatiTransactionsThesetofstudents’coursescoresintheexampleMinsup=1.5,Minconf=0.7

CaseOOPDatabaseStatisticsDatastructureMIS186778671682618789778038489867989473867984625708587727966567866187771877571808866964848897565868679108368658589TransactionsThesetofstudentMembershipFunctionsMembershipvalue

LowMiddleHigh

1

59636973788590100MembershipFunctionsMembershipSTEP1

Transformthequantitativevaluesofeachtransactiondatumintofuzzysets.e.g.Score=86LowMiddleHigh

1

7885STEP1

TransformthequantitatSTEP2CaseOOPDatabaseStatisticsDataStructureMISLMHLMHLMHLMHLMH10.00.00.70.00.70.00.00.00.70.00.80.020.80.00.00.00.00.80.00.00.90.00.70.00.00.00.90.00.00.70.00.00.940.01.00.00.00.00.70.00.00.050.00.70.00.00.00.60.00.00.80.00.90.00.00.20.00.00.00.00.00.00.870.00.80.00.00.00.80.00.80.00.00.80.00.00.40.280.00.00.70.00.60.00.00.00.890.00.80.00.00.00.70.00.00.70.00.50.1100.00.50.00.00.00.60.00.00.9countCalculatethescalarcardinalityofeachattributeregioninthetransactionsasthecountvalue

STEP2CaseOOPDatabaseStatistiSTEP3Checking

count

minimumsupportvalue(1.5)ThecontentsofL1forthisexampleItemsetSupportOOP.Middle3.8OOP.High2.3DB.Middle2.4DB.High3.8ST.Middle1.6ST.High4.6DS.Middle3.9DS.High2.5MIS.Middle2.3MIS.High4.0STEP3CheckingcountminimAnExampleSTEP4Setr=1STEP5GeneratethecandidatesetCr+1

fromLr

e.g.(OOP.Middle,DB.Middle),(OOP.Middle,DB.High)(OOP.Middle,ST.Middle),(OOP.Middle,ST.High)(OOP.Middle,DS.Middle),(OOP.Middle,DS.High)(OOP.Middle,MIS.Middle),(OOP.Middle,MIS.High)…,(DS.High,MIS.Middle),and(DS.High,MIS.High)AnExampleSTEP4STEP6DothefollowingsubstepsforeachnewlyformedcandidateitemsetStep6.1Calculatethefuzzymembershipvalueofeachtransactiondatume.g.

CaseOOP.MiddleDB.HighOOP.Middle

DB.High10.00.00.020.00.80.041.00.70.760.20.00.080.00.00.090.80.00.0100.20.00.0STEP6CaseOOP.MiddleDB.HighOOPSTEP6Step6.2Calculatethescalarcardinality(count)ofeachcandidate2-itemsetinthetransactiondataItemsetSupportOOP.MiddleDB.Middle0.6OOP.MiddleDB.High2.2OOP.MiddleST.Middle1.5OOP.MiddleST.High1.8OOP.MiddleDS.Middle1.7OOP.MiddleDS.High1.6OOP.MiddleMIS.Middle1.4OOP.MiddleMIS.High0.9……DS.HighMIS.High1.3STEP6Step6.2ItemsetSupportOOSTEP6Step6.3Checkwhetherthesecountsarelargerthanorequaltothepredefinedminimumsupportvalue1.5Result(OOP.Middle,DB.High),(OOP.Middle,ST.Middle)(OOP.Middle,ST.High),(OOP.Middle,DS.Middle)(OOP.Middle,DS.High),(OOP.High,DB.Middle)(OOP.High,MIS.High),(DB.Middle,MIS.High)(DB.High,ST.High),(DB.High,DS.Middle)(ST.High,DS.Middle),(ST.High,MIS.Middle)(ST.High,MIS.High),and(DS.Middle,MIS.Middle)STEP6Step6.3STEP7IFLr+1

isnull,thendothenextstep;otherwise,setr=r+1andrepeatSTEPs5to7e.g:TheitemsetsandtheirsupportvaluesinL3

Itemsetsupport(OOP.Middle,DB.High,andDS.Middle)1.6(DB.High,ST.High,andDS.Middle)1.9STEP7IFLr+1isnull,thendoSTEP8ConstructtheassociationrulesforeachlargeitemsetusingthefollowingsubstepsStep8.1Formallpossibleassociationrulese.g:(OOP.Middle,DB.High,DS.Middle)IfOOP.MiddleandDB.HighthenDS.Middle.IfOOP.MiddleandDS.MiddlethenDB.High.IfDB.HighandDS.MiddlethenOOP.Middle.STEP8ConstructtheassociatioSTEP8Step8.2Calculatetheconfidencefactorsfortheaboveassociationrulese.g.RuleIfOOP.MiddleandDB.HighthenDS.MiddleThecountsofOOP.MiddleDB.HighOOP.MiddleDB.HighDS.MiddleSTEP8Step8.2ConfidenceCaseOOP.MiddleDB.HighOOP.MiddleDB.HighDS.MiddleOOP.MiddleDB.HighDS.Middle10.00.00.00.80.020.00.80.00.70.060.20.00.00.00.00.80.880.00.00.00.10.090.80.00.00.00.0100.20.00.00.00.0count3.91.6==0.73ConfidenceCaseOOP.MiddleDB.HigSTEP9CheckwhethertheconfidencefactorsoftheaboveassociationrulesarelargerthanorequaltothepredefinedconfidencethresholdSTEP9CheckwhethertheconfidFuzzyassociationrules12rulesIf"OOP.MiddleandDS.Middle"then"DB.High“,conf=0.94If"ST.Middle"then"OOP.Middle“,conf=0.94If"DB.HighandST.High"then"DS.Middle",conf=0.86If"MIS.Middle"then"ST.High",conf=0.83If"DS.Middle"then"MIS.Middle“,conf=0.78If"OOP.High"then"DB.Middle“,conf=0.74If"OOP.MiddleandDB.High"then"DS.Middle“,conf=0.73If"DS.Middle"then"ST.High“,conf=0.72If"DB.Middle"then"OOP.High",conf=0.71If"DB.High"then"DS.Middle“,conf=0.71If"DB.HighandDS.Middle"then"ST.High“,conf=0.70If"OOP.High"then"MIS.High“,conf=0.70

FuzzyassociationrulesAnotherVariantUseonlytheregionwiththemaximumfuzzycardinalityforeachitemCaseOOPDatabaseStatisticsDataStructureMISLMHLMHLMHLMHLMH10.00.00.70.00.70.00.00.00.70.00.80.020.80.00.00.00.00.80.00.00.90.00.70.00.00.00.90.00.00.70.00.00.940.01.00.00.00.00.70.00.00.050.00.70.00.00.00.60.00.00.80.00.90.00.00.20.00.00.00.00.00.00.870.00.80.00.00.00.80.00.80.00.00.80.00.00.40.280.00.00.70.00.60.00.00.00.890.00.80.00.00.00.70.00.00.70.00.50.1100.00.50.00.00.00.60.00.00.9countAnotherVariantUseonlythereResultsItemsetOOP.MiddleDB.HighST.HighDS.MiddleMIS.HighL1Itemset(OOP.Middle,DB.High)(OOP.Middle,ST.High)(OOP.Middle,DS.Middle)(DB.High,ST.High)(DB.High,DS.Middle)(ST.High,DS.Middle)(ST.High,MIS.High)L2ResultsItemsetOOP.MiddleDB.HigResultsLessnumberoflargeitemsetsNotcompleteLesscomputationtimeItemset(OOP.Middle,DB.High,DS.Middle)(DB.High,ST.High,DS.Middle)L3ResultsLessnumberoflargeitTaxonomyOnlytheterminalitemscanappearintransactiondataTaxonomyOnlytheterminalitemMiningundertaxonomyManyApproachesHere,byAgrawalandSrikant’sapproachStep1:Defineminsupandminconfex:minsup=50% minconf=50%Step2:GenerateexpandedtransactiondataStep3:FindlargeitemsetsStep4:GenerateassociationrulesStep5:CheckinterestofassociationrulesMiningundertaxonomyManyApprStep2Generateexpandedtransactionancestorsareadded.TIDExpandedItems1A,C,E,T1,T2,,T32B,C,D,T1,T2,T33B,C,E,T1,T2,T34C,E,T2,T3

5A,D,T1,T2,T3

6B,C,D,T1,T2,T3

T3DET2T1CABTIDPurchaseditems1A,C,E2B,C,D3B,C,E4A,E,5A,D,6B,C,DStep2GenerateexpandedtransStep5

RuleInterestCriteria:1.Arulewithnoancestorrulesminedout,2.ThesupportvalueofarulebeingR-timelargerthantheexpectedsupportvaluesofitsancestorrules3.TheconfidencevalueofarulebeingR-timelargerthantheexpectedconfidencevaluesofitsancestorrules.Step5RuleInterestCriteria:Algorithm

InputQuantitativetransactiondatasetMembershipfunctionsTaxonomyThreeThresholdsMin-supportMin-confidenceR-interestOutputFuzzyinterestinggeneralizedassociationrulesAlgorithmInputExampleQuantitativetransactiondatasetTransactionIDSomepurchaseditemswithquantitiesTaxonomyOnlytheterminalitemscanappearintransactiondataT2T1CABT3DETIDPurchaseditems1(A,3)(C,4)(E,2)2(B,3)(C,7)(D,7)3(B,2)(C,10)(E,5)4(A,9)(E,10)5(A,7)(D,8)6(B,2)(C,8)(D,10)TaxonomyExampleQuantitativetransactioStep1GenerateexpandedtransactionAncestorsareappendedQuantitiesareaddedT3DET2T1CABTIDPurchaseditems1(A,3)(C,4)(E,2)2(B,3)(C,7)(D,7)3(B,2)(C,10)(E,5)4(A,9)(E,10)5(A,7)(D,8)6(B,2)(C,8)(D,10)TIDExpandedItems1(A,3)(C,4)(E,2)(T1,3)(T2,7)(T3,2)2(B,3)(C,7)(D,7)(T1,3

)(T2,10)(T3,7)3(B,2)(C,10)(E,5)(T1,2)(T2,12)(T3,5)4(C,9)(E,10)(T2,9)(T3,10)5(A,7)(D,8)(T1,7)(T2,7)(T3,8)6(B,2)(C,8)(D,10)(T1,2)(T2,10)(T3,10)Step1GenerateexpandedtransStep2TransformthequantitativedataintofuzzydataUsingthegivenmembershipfunctionsTakethefirstitemintransaction4asanexample

TIDExpandedItems1(A,3)(C,4)(E,2)(T1,3)(T2,7)(T3,2)2(B,3)(C,7)(D,7)(T1,3

)(T2,10)(T3,7)3(B,2)(C,10)(E,5)(T1,2)(T2,12)(T3,5)4(C,9)(E,10)(T2,9)(T3,10)5(A,7)(D,8)(T1,7)(T2,7)(T3,8)6(B,2)(C,8)(D,10)(T1,2)(T2,10)(T3,10)016911LowMiddleHigh10MembershipvalueNumberofitemStep2TransformthequantitatResultafterstep2ThefuzzysetstransformedfromthequantitativedataTIDFuzzyset123456Resultafterstep2ThefuzzysStep3Calculatethescalarcardinalityofeachfuzzyregion

Step3CalculatethescalarcarStep4GeneratelargeitemsetsL1Assumetheminimumsupportvalueis1.5Step4GeneratelargeitemsetsStep5GeneratecandidatesandcalculatefuzzycardinalityByminimumoperatorse.g.(B.Low,C.Middle)1.4Step5GeneratecandidatesandStep6GeneratelargeitemsetsL2Assumetheminimumsupportvalueis1.5NoL3L2Step6GeneratelargeitemsetsStep7CalculateconfidenceandgeneraterulesByconditionalprobabilityconfidencevalue>=min-confthresholdAssumemin-conf=0.75TherearethreerulesinthisexampleIfC=Middle,thenT1=Low(2.0,0.77)IfT1=Middle,thenT3=Middle(1.6,0.8)IfT3=Middle,thenT2=High(2.4,0.86)IfT3=High,thenT2=High(1.8,0.82)Step7CalculateconfidenceandStep8CheckwhethertherulesareinterestingAlltheruleshasnoancestorrulesFinalResults:IfC=Middle,thenT1=Low(2.0,0.77)IfT1=Middle,thenT3=Middle(1.6,0.8)IfT3=Middle,thenT2=High(2.4,0.86)IfT3=High,thenT2=High(1.8,0.82)Step8CheckwhethertherulesAnotherVariantUseonlytheregionwiththemaximumfuzzycardinalityforeachitemAnotherVariantUseonlythereResultafterstep3LessnumberoflargeitemsetsNotcompleteLesscomputationtimeItemsetsupB.Low2.2C.Middle2.6D.Middle1.6T1.Low2.8T2.High3.6T3.Middle2.8ItemsetsupB.Low,T3.Middle1.6C.Middle,T3.Middle1.6T1.Low,T3.Middle1.6T2.High,T3.Middle2.4L1L2Resultafterstep3LessnumberExperimentsInConaPentium-IIIThreelevels,fourbranchesNumbersofpurchaseditemsarefirstrandomlygeneratedPurchaseditemsandquantitiesarethenrandomlygenerated.

ExperimentsInConaPentium-IExperimentsFor10000transactionsTherelationshipsbetweennumbersofrulesandminimumsupportvaluesandconfidencevalues

Support->RulesConfidence

->RulesExperimentsFor10000transactiExperimentsDifferentaveragenumbersofitemsintransactionsAveragenumberofitems->RulesExperimentsDifferentaveragenExperimentsAveragenumberofitems->timeExperimentsExperimentsFordifferentnumbersoftransactionsNumbersoftransactions->RulesExperimentsFordifferentnumbeExperimentsNumbersoftransactions->TimeExperimentsExperimentAcomparisonoftheproposedapproachesRulesaremorecompletebythefirstapproach

0100200300400500600700800200250300350400450500550600650700750800min-supportnumberofrulesThefirstapproachThesecondapproachExperimentAcomparisonoftheExperimentAcomparisonoftheproposedapproachesMoreTimeisspentbythefirstapproach

020406080100120140160180200250300350400450500550600650700750800min-supporttime(second)ThesecondapproachThefirstapproachExperimentAcomparisonoftheExperimentAcomparisonoffuzzyeffectsBypredictionaccuracy

Fuzzypartitionisbetterthancrisppartition01020304050607080MinimumSupportValueAccuracySecondApproachFirstApproachCrispPartitionExperimentAcomparisonoffuzzWebMiningLogdataWebMiningKnowledgeandstrategyBrowsingpatternsWebMiningLogdataWebMiWebminingweb-contentminingfocusesoninformationdiscoveryfromsourcesacrosstheWorldWideWebe.g.Miningkeywordrelationsfromwebpagesweb-usageminingemphasizesontheautomaticdiscoveryofuseraccesspatternsfromwebserverse.g.MiningpagebrowsingpatternsfromlogfilesWebminingweb-contentminingGoalProposingaweb-usageminingalgorithmFindinglinguisticbrowsingbehaviorsfromdatalogsonwebserverse.g.IfbrowsingpageAalongtimeThennextbrowsingpageBashorttimeAdoptingfuzzyconceptsinanalyzingthebrowsingtimeGoalRelatedworkWebcontentminingWebusageminingYearProposedmethodauthor1998EfficientdataminingforpathtaversalpatternsM.S.Chenetal.1999EfficientalgorithmsforpredictingrequeststowebserverE.Cohenetal.1999MiningWWWbrowsingpatternsCooleyetal.RelatedworkWebcontentminingRelatedworkMiningsequentialpatternFuzzysettheoryandfuzzydataminingYearProposedmethodauthor1995MiningsequentialpatternsAgrawaletal.1997SPADEalgorithmParthasarathyetal.YearProposedmethodauthor1965FuzzysetL.A.Zadeh1988FuzzylogicL.A.Zadeh1999AdataminingalgorithmfortransactiondatawithquantitativevaluesT.P.Hongetal.2002MininglinguisticbrowsingpatternsintheworldwidewebT.P.Hongetal.RelatedworkMiningsequentialAprioriAllAlgorithmProposedbyAgrawalandSrikantForfindingsequentialpatternsFivephasesFormcustomersequencesFindlargeitemsetsMapitemsetsintointegersFindlargesequencesFindmaximallylargesequencesAprioriAllAlgorithmProposedbAprioriAllAlgorithmProposedbyAgrawalandSrikantForfindingsequentialpatternse.g.Customer110:00BCDCustomer211:00BCCustomer115:00BCDECustomer217:00EFCustomer1:BCD->BCDECustomer2:BC->EF

=>BC->E

AprioriAllAlgorithmProposedbExampleSupport=60%ExampleSupport=60%ExampleExampleFuzzyweb-miningalgorithmInputLogdataMembershipfunctionsForconvertingbrowsingdurationsintolinguistictermsThresholdMin-supOutputFuzzybrowsingpatternsFuzzyweb-miningalgorithmInpuLogdataDateTimeClient-ipServer-ipServer-portFile-name…2001-03-0105:39:5627821B.htm…2001-03-0105:40:0827821home-bg1.jpg…2001-03-0105:40:1027821line1.gif…::::::…2001-03-0105:40:2627821E.asp…::::::…2001-03-0105:40:522821D.htm…2001-03-0105:40:532821line1.gif…::::::…::::::…2001-03-0105:41:0828821Dp.htm…::::::…2001-03-0105:48:384821closingconnection…::::::…2001-03-0105:48:532821E.htm…::::::…2001-03-0105:50:130821C.asp…::::::…2001-03-0105:53:330821closingconnection

LogdataDateTimeClient-ipServeOuralgorithmStep1Thefollowingfilenamesareselected.asp,.htm,.html,.jva,.cgiandclosingconnectionThefollowingfourfieldsarekeptdate,time,cilent-ipandfile-nameDateTimeClient-ipFile-name2001-03-0105:39:5628B.htm2001-03-0105:40:2628E.asp::::2001-03-0105:53:160C.htm2001-03-0105:53:330closingconnectionOuralgorithmStep1DateTimeCliOuralgorithmStep2Thevaluesoffieldclient-iparetransformedintocontiguousintegersforconvenenceStep3ThelogdatasortedfirstbyencodedclientIDandthenbydateandtimeDateTimeClientIDFile-name2001-03-0105:39:561B.htm2001-03-0105:40:261E.asp2001-03-0105:41:081D.htm::::2001-03-0105:51:144B.asp2001-03-0105:53:164C.htm2001-03-0105:53:334closingconnectionOuralgorithmStep2DateTimeCliOuralgorithmStep4ThetimedurationsofthewebpagesbrowsedbyeachencodedclientIDarecalculatede.g.2001/03/01,05:39:56–2001/03/01,05:40:2630secondsDateTimeClientIDFile-name2001-03-0105:39:561B.htm2001-03-0105:40:261E.aspOuralgorithmStep4DateTimeClOuralgorithmStep5ThewebpagesbrowsedbyeachclientarelistedtoformbrowsingsequenceClientIDBrowsingSequence1(B,30)(E,42)(D,98)(C,91)2(D,62)(B,31)(D,102)3(A,92)(D,64)(B,29)(C,74)4(B,20)(C,101)(E,119)(B,11)(C,42)OuralgorithmStep5ClientIDBrOuralgorithmStep6ThetimedurationsarerepresentedasfuzzysetsUsingthegivenmembershipfunctionse.g.theseconditem(C,101)inClient40207080101130ShortMiddleLong10BrowsingdurationClientIDBrowsingSequence1(B,30)(E,42)(D,98)(C,91)2(D,62)(B,31)(D,102)3(A,92)(D,64)(B,29)(C,74)4(B,20)(C,101)(E,119)(B,11)(C,42)OuralgorithmStep6020Resultsafterstep6ThefuzzysetstransformedfromthequantitativedataClientIDFuzzySets1234Resultsafterstep6ThefuzzyOuralgorithmStep7:Themaximummembershipvalueforeachregionineachsequenceisfounde.g.Client2:D.Middle:max(0.8,0.0,0.6)=0.8OuralgorithmStep7:ThemaximOuralgorithmStep8:Thesupportvalueofeachregioniscalculatede.g.D.Middle:ClientIDMembershipvalueofD.Middle10.620.831.040.0D.Middle:0.6+0.8+1.0+0.0=2.4OuralgorithmStep8:ThesuppoOuralgorithmSteps9-11:Large1-sequencesaregeneratede.g.AssumeMin-sup:1.8B.Short,C.Middle,D.MiddleSteps12-15:Largek-sequencesaregeneratedCandidate2-itemsetsB.Short,C.MiddleC.Middle,B.ShortB.Short,D.MiddleD.Middle,B.Short:OuralgorithmSteps9-11:LargeSupportvalueofcompositeregions

ThesupportvalueofeachcompositeregioniscalculatedForexample:client4(B.Short,C.Middle)

max[min(1.0,0.6),min(1.0,0.4),min(1.0,0.4)]=0.6

ClientIDMembershipvalueof