iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical...
Transcript of iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical...
![Page 1: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/1.jpg)
Sınıflandırma (Classification)
Şadi Evren ŞEKER www.SadiEvrenSEKER.com
Youtube : Bilgisayar Kavramlar
02/11/16 Data Mining: Concepts and
Techniques 1
![Page 2: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/2.jpg)
Kaynaklar
n DataMining:ConceptsandTechniques,ThirdEdition,JiaweiHan,MichelineKamber,JianPei
![Page 3: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/3.jpg)
3
Chapter 8. Classification (Sınıflandırma)
n SınıflandırmayaGiriş
n K-NNAlgoritması
n KararAğaçları(DecisionTrees)
n BayesSınıflandırmaYöntemi
n KuralTabanlıSınıflandırmalar(Rule-basedClassificaGon)
n DoğruModelinseçilmesi
n BaşarıyıarIranbazıyöntemler(EnsembleTechniques)
![Page 4: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/4.jpg)
4
Supervised vs. Unsupervised Learning (Gözetimli ve Gözetimsiz Öğrenme)
n Supervisedlearning(classificaGon)
n Supervision:Thetrainingdata(observaGons,measurements,etc.)areaccompaniedbylabelsindicaGng
theclassoftheobservaGons
n Newdataisclassifiedbasedonthetrainingset
n Unsupervisedlearning(clustering)
n Theclasslabelsoftrainingdataisunknown
n Givenasetofmeasurements,observaGons,etc.withtheaimofestablishingtheexistenceofclassesorclustersinthedata
![Page 5: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/5.jpg)
5
n ClassificaGonn predictscategoricalclasslabels(discreteornominal)n classifiesdata(constructsamodel)basedonthetrainingsetandthevalues(classlabels)inaclassifyingaXributeandusesitinclassifyingnewdata
n NumericPredicGonn modelsconGnuous-valuedfuncGons,i.e.,predictsunknownormissingvalues
n TypicalapplicaGonsn Credit/loanapproval:n Medicaldiagnosis:ifatumoriscancerousorbenignn FrauddetecGon:ifatransacGonisfraudulentn WebpagecategorizaGon:whichcategoryitis
Prediction Problems: Classification vs. Numeric Prediction
![Page 6: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/6.jpg)
6
Classification—A Two-Step Process
n ModelconstrucGon:describingasetofpredeterminedclassesn Eachtuple/sampleisassumedtobelongtoapredefinedclass,as
determinedbytheclasslabelaXributen ThesetoftuplesusedformodelconstrucGonistrainingsetn ThemodelisrepresentedasclassificaGonrules,decisiontrees,or
mathemaGcalformulaen Modelusage:forclassifyingfutureorunknownobjects
n EsGmateaccuracyofthemodeln Theknownlabeloftestsampleiscomparedwiththeclassifiedresultfromthemodel
n Accuracyrateisthepercentageoftestsetsamplesthatarecorrectlyclassifiedbythemodel
n Testsetisindependentoftrainingset(otherwiseoverfi^ng)n Iftheaccuracyisacceptable,usethemodeltoclassifynewdata
n Note:Ifthetestsetisusedtoselectmodels,itiscalledvalidaGon(test)set
![Page 7: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/7.jpg)
7
Process (1): Model Construction
Training Data
NAME RANK YEARS TENUREDMike Assistant Prof 3 noMary Assistant Prof 7 yesBill Professor 2 yesJim Associate Prof 7 yesDave Assistant Prof 6 noAnne Associate Prof 3 no
Classification Algorithms
IF rank = ‘professor’ OR years > 6 THEN tenured = ‘yes’
Classifier (Model)
![Page 8: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/8.jpg)
8
Process (2): Using the Model in Prediction
Classifier
Testing Data
NAME RANK YEARS TENUREDTom Assistant Prof 2 noMerlisa Associate Prof 7 noGeorge Professor 5 yesJoseph Assistant Prof 7 yes
Unseen Data
(Jeff, Professor, 4)
Tenured?
![Page 9: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/9.jpg)
9
Chapter 8. Classification: Basic Concepts
n ClassificaGon:BasicConcepts
n DecisionTreeInducGon
n BayesClassificaGonMethods
n Rule-BasedClassificaGon
n ModelEvaluaGonandSelecGon
n TechniquestoImproveClassificaGonAccuracy:EnsembleMethods
n Summary
![Page 10: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/10.jpg)
10
Decision Tree Induction: An Example
age?
overcast
student? credit rating?
<=30 >40
no yes yes
yes
31..40
no
fair excellent yes no
age income student credit_rating buys_computer<=30 high no fair no<=30 high no excellent no31…40 high no fair yes>40 medium no fair yes>40 low yes fair yes>40 low yes excellent no31…40 low yes excellent yes<=30 medium no fair no<=30 low yes fair yes>40 medium yes fair yes<=30 medium yes excellent yes31…40 medium no excellent yes31…40 high yes fair yes>40 medium no excellent no
q Trainingdataset:Buys_computerq ThedatasetfollowsanexampleofQuinlan’sID3(PlayingTennis)
q ResulGngtree:
![Page 11: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/11.jpg)
11
Algorithm for Decision Tree Induction n Basicalgorithm(agreedyalgorithm)
n Treeisconstructedinatop-downrecursivedivide-and-conquermanner
n Atstart,allthetrainingexamplesareattherootn AXributesarecategorical(ifconGnuous-valued,theyarediscreGzedinadvance)
n ExamplesareparGGonedrecursivelybasedonselectedaXributes
n TestaXributesareselectedonthebasisofaheurisGcorstaGsGcalmeasure(e.g.,informaGongain)
n CondiGonsforstoppingparGGoningn Allsamplesforagivennodebelongtothesameclassn TherearenoremainingaXributesforfurtherparGGoning–majorityvoGngisemployedforclassifyingtheleaf
n Therearenosamplesled
![Page 12: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/12.jpg)
Brief Review of Entropy
n
12
m = 2
![Page 13: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/13.jpg)
13
Attribute Selection Measure: Information Gain (ID3/C4.5)
n SelecttheaXributewiththehighestinformaGongainn LetpibetheprobabilitythatanarbitrarytupleinDbelongsto
classCi,esGmatedby|Ci,D|/|D|n ExpectedinformaGon(entropy)neededtoclassifyatupleinD:
n InformaGonneeded(aderusingAtosplitDintovparGGons)toclassifyD:
n InformaGongainedbybranchingonaXributeA
)(log)( 21
i
m
ii ppDInfo ∑
=
−=
)(||||
)(1
j
v
j
jA DInfo
DD
DInfo ×=∑=
(D)InfoInfo(D)Gain(A) A−=
![Page 14: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/14.jpg)
14
Attribute Selection: Information Gain
g ClassP:buys_computer=“yes”g ClassN:buys_computer=“no”
means“age<=30”has5outof14samples,with2yes’esand3no’s.Hence
Similarly,
age pi ni I(pi, ni)<=30 2 3 0.97131…40 4 0 0>40 3 2 0.971
694.0)2,3(145
)0,4(144)3,2(
145)(
=+
+=
I
IIDInfoage
048.0)_(151.0)(029.0)(
===
ratingcreditGainstudentGainincomeGain
246.0)()()( =−= DInfoDInfoageGain ageage income student credit_rating buys_computer
<=30 high no fair no<=30 high no excellent no31…40 high no fair yes>40 medium no fair yes>40 low yes fair yes>40 low yes excellent no31…40 low yes excellent yes<=30 medium no fair no<=30 low yes fair yes>40 medium yes fair yes<=30 medium yes excellent yes31…40 medium no excellent yes31…40 high yes fair yes>40 medium no excellent no
)3,2(145 I
940.0)145(log
145)
149(log
149)5,9()( 22 =−−== IDInfo
![Page 15: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/15.jpg)
15
Computing Information-Gain for Continuous-Valued Attributes
n LetaXributeAbeaconGnuous-valuedaXribute
n MustdeterminethebestsplitpointforA
n SortthevalueAinincreasingorder
n Typically,themidpointbetweeneachpairofadjacentvaluesisconsideredasapossiblesplitpoint
n (ai+ai+1)/2isthemidpointbetweenthevaluesofaiandai+1
n Thepointwiththeminimumexpectedinforma4onrequirementforAisselectedasthesplit-pointforA
n Split:
n D1isthesetoftuplesinDsaGsfyingA≤split-point,andD2isthesetoftuplesinDsaGsfyingA>split-point
![Page 16: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/16.jpg)
16
Gain Ratio for Attribute Selection (C4.5)
n InformaGongainmeasureisbiasedtowardsaXributeswithalargenumberofvalues
n C4.5(asuccessorofID3)usesgainraGotoovercometheproblem(normalizaGontoinformaGongain)
n GainRaGo(A)=Gain(A)/SplitInfo(A)n Ex.
n gain_raGo(income)=0.029/1.557=0.019n TheaXributewiththemaximumgainraGoisselectedasthe
spli^ngaXribute
)||||
(log||||
)( 21 D
DDD
DSplitInfo jv
j
jA ×−= ∑
=
![Page 17: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/17.jpg)
17
Gini Index (CART, IBM IntelligentMiner)
n IfadatasetDcontainsexamplesfromnclasses,giniindex,gini(D)isdefinedas
wherepjistherelaGvefrequencyofclassjinDn IfadatasetDissplitonAintotwosubsetsD1andD2,thegini
indexgini(D)isdefinedas
n ReducGoninImpurity:
n TheaXributeprovidesthesmallestginisplit(D)(orthelargestreducGoninimpurity)ischosentosplitthenode(needtoenumerateallthepossiblespli;ngpointsforeacha<ribute)
∑=
−=n
jp jDgini121)(
)(||||)(
||||)( 2
21
1 DginiDD
DginiDDDginiA +=
)()()( DginiDginiAgini A−=Δ
![Page 18: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/18.jpg)
18
Computation of Gini Index
n Ex.Dhas9tuplesinbuys_computer=“yes”and5in“no”
n SupposetheaXributeincomeparGGonsDinto10inD1:{low,medium}and4inD2
Gini{low,high}is0.458;Gini{medium,high}is0.450.Thus,splitonthe{low,medium}(and{high})sinceithasthelowestGiniindex
n AllaXributesareassumedconGnuous-valuedn Mayneedothertools,e.g.,clustering,togetthepossiblesplit
valuesn CanbemodifiedforcategoricalaXributes
459.0145
1491)(
22
=⎟⎠
⎞⎜⎝
⎛−⎟⎠
⎞⎜⎝
⎛−=Dgini
)(144)(
1410)( 21},{ DGiniDGiniDgini mediumlowincome ⎟
⎠
⎞⎜⎝
⎛+⎟⎠
⎞⎜⎝
⎛=∈
![Page 19: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/19.jpg)
19
Comparing Attribute Selection Measures
n Thethreemeasures,ingeneral,returngoodresultsbutn Informa/ongain:
n biasedtowardsmulGvaluedaXributesn Gainra/o:
n tendstopreferunbalancedsplitsinwhichoneparGGonismuchsmallerthantheothers
n Giniindex:n biasedtomulGvaluedaXributes
n hasdifficultywhen#ofclassesislargen tendstofavorteststhatresultinequal-sizedparGGonsandpurityinbothparGGons
![Page 20: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/20.jpg)
20
Other Attribute Selection Measures
n CHAID:apopulardecisiontreealgorithm,measurebasedonχ2testforindependence
n C-SEP:performsbeXerthaninfo.gainandginiindexincertaincases
n G-staGsGc:hasacloseapproximaGontoχ2distribuGon
n MDL(MinimalDescripGonLength)principle(i.e.,thesimplestsoluGonispreferred):
n Thebesttreeastheonethatrequiresthefewest#ofbitstoboth(1)encodethetree,and(2)encodetheexcepGonstothetree
n MulGvariatesplits(parGGonbasedonmulGplevariablecombinaGons)
n CART:findsmulGvariatesplitsbasedonalinearcomb.ofaXrs.
n WhichaXributeselecGonmeasureisthebest?
n Mostgivegoodresults,noneissignificantlysuperiorthanothers
![Page 21: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/21.jpg)
21
Overfitting and Tree Pruning
n Overfi^ng:Aninducedtreemayoverfitthetrainingdatan Toomanybranches,somemayreflectanomaliesduetonoiseoroutliers
n Pooraccuracyforunseensamplesn Twoapproachestoavoidoverfi^ng
n Prepruning:Halttreeconstruc4onearly̵donotsplitanodeifthiswouldresultinthegoodnessmeasurefallingbelowathreshold
n Difficulttochooseanappropriatethresholdn Postpruning:Removebranchesfroma“fullygrown”tree—getasequenceofprogressivelyprunedtrees
n Useasetofdatadifferentfromthetrainingdatatodecidewhichisthe“bestprunedtree”
![Page 22: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/22.jpg)
22
Enhancements to Basic Decision Tree Induction
n Allowforcon/nuous-valueda9ributesn Dynamicallydefinenewdiscrete-valuedaXributesthatparGGontheconGnuousaXributevalueintoadiscretesetofintervals
n Handlemissinga9ributevaluesn AssignthemostcommonvalueoftheaXribute
n Assignprobabilitytoeachofthepossiblevaluesn A9ributeconstruc/on
n CreatenewaXributesbasedonexisGngonesthataresparselyrepresented
n ThisreducesfragmentaGon,repeGGon,andreplicaGon
![Page 23: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/23.jpg)
23
Classification in Large Databases
n ClassificaGon—aclassicalproblemextensivelystudiedbystaGsGciansandmachinelearningresearchers
n Scalability:ClassifyingdatasetswithmillionsofexamplesandhundredsofaXributeswithreasonablespeed
n WhyisdecisiontreeinducGonpopular?n relaGvelyfasterlearningspeed(thanotherclassificaGonmethods)
n converGbletosimpleandeasytounderstandclassificaGonrules
n canuseSQLqueriesforaccessingdatabasesn comparableclassificaGonaccuracywithothermethods
n RainForest(VLDB’98—Gehrke,Ramakrishnan&GanG)n BuildsanAVC-list(aXribute,value,classlabel)
![Page 24: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/24.jpg)
24
Scalability Framework for RainForest
n Separates the scalability aspects from the criteria that determine the quality of the tree
n Builds an AVC-list: AVC (Attribute, Value, Class_label)
n AVC-set (of an attribute X )
n Projection of training dataset onto the attribute X and class label where counts of individual class label are aggregated
n AVC-group (of a node n )
n Set of AVC-sets of all predictor attributes at the node n
![Page 25: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/25.jpg)
25
Rainforest: Training Set and Its AVC Sets
student Buy_Computer
yes no
yes 6 1
no 3 4
Age Buy_Computer
yes no
<=30 2 3
31..40 4 0
>40 3 2
Credit rating
Buy_Computer
yes no
fair 6 2
excellent 3 3
age income studentcredit_ratingbuys_computer<=30 high no fair no<=30 high no excellent no31…40 high no fair yes>40 medium no fair yes>40 low yes fair yes>40 low yes excellent no31…40 low yes excellent yes<=30 medium no fair no<=30 low yes fair yes>40 medium yes fair yes<=30 medium yes excellent yes31…40 medium no excellent yes31…40 high yes fair yes>40 medium no excellent no
AVC-set on income AVC-set on Age
AVC-set on Student
Training Examples income Buy_Computer
yes no
high 2 2
medium 4 2
low 3 1
AVC-set on credit_rating
![Page 26: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/26.jpg)
26
BOAT (Bootstrapped Optimistic Algorithm for Tree Construction)
n Use a statistical technique called bootstrapping to create several smaller samples (subsets), each fits in memory
n Each subset is used to create a tree, resulting in several trees
n These trees are examined and used to construct a new tree T’
n It turns out that T’ is very close to the tree that would be generated using the whole data set together
n Adv: requires only two scans of DB, an incremental alg.
![Page 27: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/27.jpg)
November 2, 2016 Data Mining: Concepts and Techniques 27
Presentation of Classification Results
![Page 28: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/28.jpg)
November 2, 2016 Data Mining: Concepts and Techniques 28
Visualization of a Decision Tree in SGI/MineSet 3.0
![Page 29: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/29.jpg)
Data Mining: Concepts and Techniques 29
Interactive Visual Mining by Perception-Based Classification (PBC)
![Page 30: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/30.jpg)
30
Chapter 8. Classification: Basic Concepts
n ClassificaGon:BasicConcepts
n DecisionTreeInducGon
n BayesClassificaGonMethods
n Rule-BasedClassificaGon
n ModelEvaluaGonandSelecGon
n TechniquestoImproveClassificaGonAccuracy:EnsembleMethods
n Summary
![Page 31: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/31.jpg)
31
Bayesian Classification: Why?
n AstaGsGcalclassifier:performsprobabilis4cpredic4on,i.e.,predictsclassmembershipprobabiliGes
n FoundaGon:BasedonBayes’Theorem.n Performance:AsimpleBayesianclassifier,naïveBayesian
classifier,hascomparableperformancewithdecisiontreeandselectedneuralnetworkclassifiers
n Incremental:Eachtrainingexamplecanincrementallyincrease/decreasetheprobabilitythatahypothesisiscorrect—priorknowledgecanbecombinedwithobserveddata
n Standard:EvenwhenBayesianmethodsarecomputaGonallyintractable,theycanprovideastandardofopGmaldecisionmakingagainstwhichothermethodscanbemeasured
![Page 32: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/32.jpg)
32
Bayes’ Theorem: Basics n TotalprobabilityTheorem:
n Bayes’Theorem:
n LetXbeadatasample(“evidence”):classlabelisunknownn LetHbeahypothesisthatXbelongstoclassCn ClassificaGonistodetermineP(H|X),(i.e.,posterioriprobability):the
probabilitythatthehypothesisholdsgiventheobserveddatasampleXn P(H)(priorprobability):theiniGalprobability
n E.g.,Xwillbuycomputer,regardlessofage,income,…n P(X):probabilitythatsampledataisobservedn P(X|H)(likelihood):theprobabilityofobservingthesampleX,giventhat
thehypothesisholdsn E.g.,GiventhatXwillbuycomputer,theprob.thatXis31..40,mediumincome
)()1
|()( iAPM
i iABPBP ∑
==
)(/)()|()()()|()|( XXX
XX PHPHPPHPHPHP ×==
![Page 33: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/33.jpg)
33
Prediction Based on Bayes’ Theorem
n GiventrainingdataX,posterioriprobabilityofahypothesisH,P(H|X),followstheBayes’theorem
n Informally,thiscanbeviewedas
posteriori=likelihoodxprior/evidence
n PredictsXbelongstoCiifftheprobabilityP(Ci|X)isthehighestamongalltheP(Ck|X)forallthekclasses
n PracGcaldifficulty:ItrequiresiniGalknowledgeofmanyprobabiliGes,involvingsignificantcomputaGonalcost
)(/)()|()()()|()|( XXX
XX PHPHPPHPHPHP ×==
![Page 34: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/34.jpg)
34
Classification Is to Derive the Maximum Posteriori
n LetDbeatrainingsetoftuplesandtheirassociatedclasslabels,andeachtupleisrepresentedbyann-DaXributevectorX=(x1,x2,…,xn)
n SupposetherearemclassesC1,C2,…,Cm.n ClassificaGonistoderivethemaximumposteriori,i.e.,the
maximalP(Ci|X)n ThiscanbederivedfromBayes’theorem
n SinceP(X)isconstantforallclasses,only
needstobemaximized
)()()|(
)|( XX
X PiCPiCP
iCP =
)()|()|( iCPiCPiCP XX =
![Page 35: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/35.jpg)
35
Naïve Bayes Classifier
n AsimplifiedassumpGon:aXributesarecondiGonallyindependent(i.e.,nodependencerelaGonbetweenaXributes):
n ThisgreatlyreducesthecomputaGoncost:OnlycountstheclassdistribuGon
n IfAkiscategorical,P(xk|Ci)isthe#oftuplesinCihavingvaluexkforAkdividedby|Ci,D|(#oftuplesofCiinD)
n IfAkisconGnous-valued,P(xk|Ci)isusuallycomputedbasedonGaussiandistribuGonwithameanμandstandarddeviaGonσ
andP(xk|Ci)is
)|(...)|()|(1
)|()|(21
CixPCixPCixPn
kCixPCiP
nk×××=∏
==X
2
2
2)(
21),,( σ
µ
σπσµ
−−
=x
exg
),,()|(ii CCkxgCiP σµ=X
![Page 36: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/36.jpg)
36
Naïve Bayes Classifier: Training Dataset
Class:C1:buys_computer=‘yes’C2:buys_computer=‘no’Datatobeclassified:X=(age<=30,Income=medium,Student=yesCredit_raGng=Fair)
age income studentcredit_ratingbuys_computer<=30 high no fair no<=30 high no excellent no31…40 high no fair yes>40 medium no fair yes>40 low yes fair yes>40 low yes excellent no31…40 low yes excellent yes<=30 medium no fair no<=30 low yes fair yes>40 medium yes fair yes<=30 medium yes excellent yes31…40 medium no excellent yes31…40 high yes fair yes>40 medium no excellent no
![Page 37: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/37.jpg)
37
Naïve Bayes Classifier: An Example n P(Ci):P(buys_computer=“yes”)=9/14=0.643P(buys_computer=“no”)=5/14=0.357n ComputeP(X|Ci)foreachclass
P(age=“<=30”|buys_computer=“yes”)=2/9=0.222P(age=“<=30”|buys_computer=“no”)=3/5=0.6P(income=“medium”|buys_computer=“yes”)=4/9=0.444P(income=“medium”|buys_computer=“no”)=2/5=0.4P(student=“yes”|buys_computer=“yes)=6/9=0.667P(student=“yes”|buys_computer=“no”)=1/5=0.2P(credit_raGng=“fair”|buys_computer=“yes”)=6/9=0.667P(credit_raGng=“fair”|buys_computer=“no”)=2/5=0.4
n X=(age<=30,income=medium,student=yes,credit_ra/ng=fair)P(X|Ci):P(X|buys_computer=“yes”)=0.222x0.444x0.667x0.667=0.044P(X|buys_computer=“no”)=0.6x0.4x0.2x0.4=0.019P(X|Ci)*P(Ci):P(X|buys_computer=“yes”)*P(buys_computer=“yes”)=0.028
P(X|buys_computer=“no”)*P(buys_computer=“no”)=0.007Therefore,Xbelongstoclass(“buys_computer=yes”)
age income studentcredit_ratingbuys_computer<=30 high no fair no<=30 high no excellent no31…40 high no fair yes>40 medium no fair yes>40 low yes fair yes>40 low yes excellent no31…40 low yes excellent yes<=30 medium no fair no<=30 low yes fair yes>40 medium yes fair yes<=30 medium yes excellent yes31…40 medium no excellent yes31…40 high yes fair yes>40 medium no excellent no
![Page 38: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/38.jpg)
38
Avoiding the Zero-Probability Problem
n NaïveBayesianpredicGonrequireseachcondiGonalprob.benon-zero.Otherwise,thepredictedprob.willbezero
n Ex.Supposeadatasetwith1000tuples,income=low(0),
income=medium(990),andincome=high(10)n UseLaplaciancorrec/on(orLaplacianesGmator)
n Adding1toeachcaseProb(income=low)=1/1003Prob(income=medium)=991/1003Prob(income=high)=11/1003
n The“corrected”prob.esGmatesareclosetotheir“uncorrected”counterparts
∏=
=n
kCixkPCiXP
1)|()|(
![Page 39: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/39.jpg)
39
Naïve Bayes Classifier: Comments
n Advantagesn Easytoimplementn Goodresultsobtainedinmostofthecases
n Disadvantagesn AssumpGon:classcondiGonalindependence,thereforelossofaccuracy
n PracGcally,dependenciesexistamongvariablesn E.g.,hospitals:paGents:Profile:age,familyhistory,etc.
Symptoms:fever,coughetc.,Disease:lungcancer,diabetes,etc.
n DependenciesamongthesecannotbemodeledbyNaïveBayesClassifier
n Howtodealwiththesedependencies?BayesianBeliefNetworks(Chapter9)
![Page 40: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/40.jpg)
40
Chapter 8. Classification: Basic Concepts
n ClassificaGon:BasicConcepts
n DecisionTreeInducGon
n BayesClassificaGonMethods
n Rule-BasedClassificaGon
n ModelEvaluaGonandSelecGon
n TechniquestoImproveClassificaGonAccuracy:EnsembleMethods
n Summary
![Page 41: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/41.jpg)
41
Using IF-THEN Rules for Classification
n RepresenttheknowledgeintheformofIF-THENrules
R:IFage=youthANDstudent=yesTHENbuys_computer=yesn Ruleantecedent/precondiGonvs.ruleconsequent
n Assessmentofarule:coverageandaccuracyn ncovers=#oftuplescoveredbyRn ncorrect=#oftuplescorrectlyclassifiedbyRcoverage(R)=ncovers/|D|/*D:trainingdataset*/accuracy(R)=ncorrect/ncovers
n Ifmorethanonerulearetriggered,needconflictresolu/onn Sizeordering:assignthehighestprioritytothetriggeringrulesthathas
the“toughest”requirement(i.e.,withthemosta<ributetests)n Class-basedordering:decreasingorderofprevalenceormisclassifica4on
costperclassn Rule-basedordering(decisionlist):rulesareorganizedintoonelong
prioritylist,accordingtosomemeasureofrulequalityorbyexperts
![Page 42: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/42.jpg)
42
age?
student? credit rating?
<=30 >40
no yes yes
yes
31..40
no
fair excellent yes no
n Example:RuleextracGonfromourbuys_computerdecision-treeIFage=youngANDstudent=noTHENbuys_computer=noIFage=youngANDstudent=yesTHENbuys_computer=yesIFage=mid-age THENbuys_computer=yesIFage=oldANDcredit_ra4ng=excellentTHENbuys_computer=noIFage=oldANDcredit_ra4ng=fairTHENbuys_computer=yes
Rule Extraction from a Decision Tree n Rulesareeasiertounderstandthanlarge
treesn Oneruleiscreatedforeachpathfromthe
roottoaleafn EachaXribute-valuepairalongapathformsa
conjuncGon:theleafholdstheclasspredicGon
n RulesaremutuallyexclusiveandexhausGve
![Page 43: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/43.jpg)
43
Rule Induction: Sequential Covering Method
n SequenGalcoveringalgorithm:Extractsrulesdirectlyfromtrainingdata
n TypicalsequenGalcoveringalgorithms:FOIL,AQ,CN2,RIPPERn Rulesarelearnedsequen4ally,eachforagivenclassCiwillcover
manytuplesofCibutnone(orfew)ofthetuplesofotherclassesn Steps:
n RulesarelearnedoneataGmen EachGmearuleislearned,thetuplescoveredbytherulesareremoved
n RepeattheprocessontheremainingtuplesunGltermina4oncondi4on,e.g.,whennomoretrainingexamplesorwhenthequalityofarulereturnedisbelowauser-specifiedthreshold
n Comp.w.decision-treeinducGon:learningasetofrulessimultaneously
![Page 44: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/44.jpg)
44
Sequential Covering Algorithm
while(enoughtargettuplesled)generatearuleremoveposiGvetargettuplessaGsfyingthisrule
Examples covered by Rule 3
Examples covered by Rule 2 Examples covered
by Rule 1
Positive examples
![Page 45: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/45.jpg)
45
Rule Generation n Togeneratearule
while(true)findthebestpredicatepiffoil-gain(p)>thresholdthenaddptocurrentruleelsebreak
Positive examples
Negative examples
A3=1 A3=1&&A1=2 A3=1&&A1=2
&&A8=5
![Page 46: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/46.jpg)
46
How to Learn-One-Rule? n Startwiththemostgeneralrulepossible:condiGon=emptyn Addingnewa<ributesbyadopGngagreedydepth-firststrategy
n Pickstheonethatmostimprovestherulequalityn Rule-Qualitymeasures:considerbothcoverageandaccuracy
n Foil-gain(inFOIL&RIPPER):assessesinfo_gainbyextendingcondiGon
n favorsrulesthathavehighaccuracyandcovermanyposiGvetuples
n Rulepruningbasedonanindependentsetoftesttuples
Pos/negare#ofposiGve/negaGvetuplescoveredbyR.IfFOIL_PruneishigherfortheprunedversionofR,pruneR
)log''
'(log'_ 22 negpospos
negposposposGainFOIL
+−
+×=
negposnegposRPruneFOIL
+
−=)(_
![Page 47: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/47.jpg)
47
Chapter 8. Classification: Basic Concepts
n ClassificaGon:BasicConcepts
n DecisionTreeInducGon
n BayesClassificaGonMethods
n Rule-BasedClassificaGon
n ModelEvaluaGonandSelecGon
n TechniquestoImproveClassificaGonAccuracy:EnsembleMethods
n Summary
![Page 48: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/48.jpg)
Model Evaluation and Selection
n EvaluaGonmetrics:Howcanwemeasureaccuracy?Othermetricstoconsider?
n Usevalida/ontestsetofclass-labeledtuplesinsteadoftrainingsetwhenassessingaccuracy
n MethodsforesGmaGngaclassifier’saccuracy:n Holdoutmethod,randomsubsampling
n Cross-validaGonn Bootstrap
n Comparingclassifiers:n Confidenceintervals
n Cost-benefitanalysisandROCCurves
48
![Page 49: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/49.jpg)
Classifier Evaluation Metrics: Confusion Matrix
Actualclass\Predictedclass
buy_computer=yes
buy_computer=no
Total
buy_computer=yes 6954 46 7000buy_computer=no 412 2588 3000
Total 7366 2634 10000
n Givenmclasses,anentry,CMi,jinaconfusionmatrixindicates#oftuplesinclassithatwerelabeledbytheclassifierasclassj
n Mayhaveextrarows/columnstoprovidetotals
ConfusionMatrix:Actualclass\Predictedclass C1 ¬C1
C1 TruePosi/ves(TP) FalseNega/ves(FN)
¬C1 FalsePosi/ves(FP) TrueNega/ves(TN)
Example of Confusion Matrix:
49
![Page 50: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/50.jpg)
Classifier Evaluation Metrics: Accuracy, Error Rate, Sensitivity and
Specificity
n ClassifierAccuracy,orrecogniGonrate:percentageoftestsettuplesthatarecorrectlyclassifiedAccuracy=(TP+TN)/All
n Errorrate:1–accuracy,orErrorrate=(FP+FN)/All
n ClassImbalanceProblem:n Oneclassmayberare,e.g.fraud,orHIV-posiGve
n Significantmajorityofthenega4veclassandminorityoftheposiGveclass
n Sensi/vity:TruePosiGverecogniGonrate
n Sensi/vity=TP/Pn Specificity:TrueNegaGverecogniGonrate
n Specificity=TN/N
A\P C ¬C
C TP FN P
¬C FP TN N
P’ N’ All
50
![Page 51: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/51.jpg)
Classifier Evaluation Metrics: Precision and Recall, and F-
measures n Precision:exactness–what%oftuplesthattheclassifier
labeledasposiGveareactuallyposiGve
n Recall:completeness–what%ofposiGvetuplesdidtheclassifierlabelasposiGve?
n Perfectscoreis1.0n InverserelaGonshipbetweenprecision&recalln Fmeasure(F1orF-score):harmonicmeanofprecisionand
recall,n Fß:weightedmeasureofprecisionandrecall
n assignsßGmesasmuchweighttorecallastoprecision
51
![Page 52: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/52.jpg)
Classifier Evaluation Metrics: Example
52
n Precision=90/230=39.13%Recall=90/300=30.00%
ActualClass\Predictedclass cancer=yes cancer=no Total RecogniGon(%)
cancer=yes 90 210 300 30.00(sensi4vity
cancer=no 140 9560 9700 98.56(specificity)
Total 230 9770 10000 96.40(accuracy)
![Page 53: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/53.jpg)
Evaluating Classifier Accuracy: Holdout & Cross-Validation
Methods n Holdoutmethod
n GivendataisrandomlyparGGonedintotwoindependentsetsn Trainingset(e.g.,2/3)formodelconstrucGonn Testset(e.g.,1/3)foraccuracyesGmaGon
n Randomsampling:avariaGonofholdoutn RepeatholdoutkGmes,accuracy=avg.oftheaccuraciesobtained
n Cross-valida/on(k-fold,wherek=10ismostpopular)n RandomlyparGGonthedataintokmutuallyexclusivesubsets,eachapproximatelyequalsize
n Ati-thiteraGon,useDiastestsetandothersastrainingsetn Leave-one-out:kfoldswherek=#oftuples,forsmallsizeddata
n *Stra/fiedcross-valida/on*:foldsarestraGfiedsothatclassdist.ineachfoldisapprox.thesameasthatintheiniGaldata
53
![Page 54: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/54.jpg)
Evaluating Classifier Accuracy: Bootstrap
n Bootstrapn Workswellwithsmalldatasetsn Samplesthegiventrainingtuplesuniformlywithreplacement
n i.e.,eachGmeatupleisselected,itisequallylikelytobeselectedagainandre-addedtothetrainingset
n Severalbootstrapmethods,andacommononeis.632boostrapn AdatasetwithdtuplesissampleddGmes,withreplacement,resulGngin
atrainingsetofdsamples.Thedatatuplesthatdidnotmakeitintothetrainingsetendupformingthetestset.About63.2%oftheoriginaldataendupinthebootstrap,andtheremaining36.8%formthetestset(since(1–1/d)d≈e-1=0.368)
n RepeatthesamplingprocedurekGmes,overallaccuracyofthemodel:
54
![Page 55: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/55.jpg)
Estimating Confidence Intervals: Classifier Models M1 vs. M2
n Supposewehave2classifiers,M1andM2,whichoneisbeXer?
n Use10-foldcross-validaGontoobtainand
n Thesemeanerrorratesarejustes4matesoferroronthetrue
populaGonoffuturedatacases
n Whatifthedifferencebetweenthe2errorratesisjust
aXributedtochance?
n Useatestofsta/s/calsignificance
n ObtainconfidencelimitsforourerroresGmates
55
![Page 56: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/56.jpg)
Estimating Confidence Intervals: Null Hypothesis
n Perform10-foldcross-validaGon
n Assumesamplesfollowatdistribu/onwithk–1degreesoffreedom(here,k=10)
n Uset-test(orStudent’st-test)
n NullHypothesis:M1&M2arethesame
n Ifwecanrejectnullhypothesis,then
n weconcludethatthedifferencebetweenM1&M2issta/s/callysignificant
n Chosemodelwithlowererrorrate
56
![Page 57: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/57.jpg)
Estimating Confidence Intervals: t-test
n Ifonly1testsetavailable:pairwisecomparisonn Forithroundof10-foldcross-validaGon,thesamecrossparGGoningisusedtoobtainerr(M1)ianderr(M2)i
n Averageover10roundstoget
n t-testcomputest-sta/s/cwithk-1degreesoffreedom:
n Iftwotestsetsavailable:usenon-pairedt-test
where
and
where
where k1 & k2 are # of cross-validation samples used for M1 & M2, resp. 57
![Page 58: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/58.jpg)
Estimating Confidence Intervals: Table for t-distribution
n Symmetricn Significancelevel,
e.g.,sig=0.05or5%meansM1&M2aresignificantlydifferentfor95%ofpopulaGon
n Confidencelimit,z=sig/2
58
![Page 59: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/59.jpg)
Estimating Confidence Intervals: Statistical Significance
n AreM1&M2significantlydifferent?n Computet.Selectsignificancelevel(e.g.sig=5%)n Consulttablefort-distribuGon:Findtvaluecorrespondingtok-1degreesoffreedom(here,9)
n t-distribuGonissymmetric:typicallyupper%pointsofdistribuGonshown→lookupvalueforconfidencelimitz=sig/2(here,0.025)
n Ift>zort<-z,thentvalueliesinrejecGonregion:n RejectnullhypothesisthatmeanerrorratesofM1&M2aresame
n Conclude:staGsGcallysignificantdifferencebetweenM1&M2
n Otherwise,concludethatanydifferenceischance59
![Page 60: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/60.jpg)
Model Selection: ROC Curves
n ROC(ReceiverOperaGngCharacterisGcs)curves:forvisualcomparisonofclassificaGonmodels
n OriginatedfromsignaldetecGontheoryn Showsthetrade-offbetweenthetrue
posiGverateandthefalseposiGveraten TheareaundertheROCcurveisa
measureoftheaccuracyofthemodeln Rankthetesttuplesindecreasing
order:theonethatismostlikelytobelongtotheposiGveclassappearsatthetopofthelist
n Theclosertothediagonalline(i.e.,theclosertheareaisto0.5),thelessaccurateisthemodel
n VerGcalaxisrepresentsthetrueposiGverate
n Horizontalaxisrep.thefalseposiGverate
n Theplotalsoshowsadiagonalline
n Amodelwithperfectaccuracywillhaveanareaof1.0
60
![Page 61: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/61.jpg)
Issues Affecting Model Selection
n Accuracyn classifieraccuracy:predicGngclasslabel
n Speedn Gmetoconstructthemodel(trainingGme)
n Gmetousethemodel(classificaGon/predicGonGme)n Robustness:handlingnoiseandmissingvaluesn Scalability:efficiencyindisk-residentdatabases
n Interpretabilityn understandingandinsightprovidedbythemodel
n Othermeasures,e.g.,goodnessofrules,suchasdecisiontreesizeorcompactnessofclassificaGonrules
61
![Page 62: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/62.jpg)
62
Chapter 8. Classification: Basic Concepts
n ClassificaGon:BasicConcepts
n DecisionTreeInducGon
n BayesClassificaGonMethods
n Rule-BasedClassificaGon
n ModelEvaluaGonandSelecGon
n TechniquestoImproveClassificaGonAccuracy:EnsembleMethods
n Summary
![Page 63: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/63.jpg)
Ensemble Methods: Increasing the Accuracy
n Ensemblemethodsn UseacombinaGonofmodelstoincreaseaccuracyn Combineaseriesofklearnedmodels,M1,M2,…,Mk,withtheaimofcreaGnganimprovedmodelM*
n Popularensemblemethodsn Bagging:averagingthepredicGonoveracollecGonofclassifiers
n BoosGng:weightedvotewithacollecGonofclassifiersn Ensemble:combiningasetofheterogeneousclassifiers
63
![Page 64: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/64.jpg)
Bagging: Boostrap Aggregation
n Analogy:DiagnosisbasedonmulGpledoctors’majorityvoten Training
n GivenasetDofdtuples,ateachiteraGoni,atrainingsetDiofdtuplesissampledwithreplacementfromD(i.e.,bootstrap)
n AclassifiermodelMiislearnedforeachtrainingsetDin ClassificaGon:classifyanunknownsampleX
n EachclassifierMireturnsitsclasspredicGonn ThebaggedclassifierM*countsthevotesandassignstheclasswiththe
mostvotestoXn PredicGon:canbeappliedtothepredicGonofconGnuousvaluesbytaking
theaveragevalueofeachpredicGonforagiventesttuplen Accuracy
n OdensignificantlybeXerthanasingleclassifierderivedfromDn Fornoisedata:notconsiderablyworse,morerobustn ProvedimprovedaccuracyinpredicGon
64
![Page 65: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/65.jpg)
Boosting
n Analogy:Consultseveraldoctors,basedonacombinaGonofweighteddiagnoses—weightassignedbasedonthepreviousdiagnosisaccuracy
n HowboosGngworks?n Weightsareassignedtoeachtrainingtuplen AseriesofkclassifiersisiteraGvelylearnedn AderaclassifierMiislearned,theweightsareupdatedto
allowthesubsequentclassifier,Mi+1,topaymorea9en/ontothetrainingtuplesthatweremisclassifiedbyMi
n ThefinalM*combinesthevotesofeachindividualclassifier,wheretheweightofeachclassifier'svoteisafuncGonofitsaccuracy
n BoosGngalgorithmcanbeextendedfornumericpredicGonn Comparingwithbagging:BoosGngtendstohavegreateraccuracy,
butitalsorisksoverfi^ngthemodeltomisclassifieddata65
![Page 66: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/66.jpg)
66
Adaboost (Freund and Schapire, 1997)
n Givenasetofdclass-labeledtuples,(X1,y1),…,(Xd,yd)n IniGally,alltheweightsoftuplesaresetthesame(1/d)n Generatekclassifiersinkrounds.Atroundi,
n TuplesfromDaresampled(withreplacement)toformatrainingsetDiofthesamesize
n Eachtuple’schanceofbeingselectedisbasedonitsweightn AclassificaGonmodelMiisderivedfromDi
n ItserrorrateiscalculatedusingDiasatestsetn Ifatupleismisclassified,itsweightisincreased,o.w.itisdecreased
n Errorrate:err(Xj)isthemisclassificaGonerroroftupleXj.ClassifierMierrorrateisthesumoftheweightsofthemisclassifiedtuples:
n TheweightofclassifierMi’svoteis
)()(1log
i
i
MerrorMerror−
∑ ×=d
jji errwMerror )()( jX
![Page 67: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/67.jpg)
Random Forest (Breiman 2001) n RandomForest:
n EachclassifierintheensembleisadecisiontreeclassifierandisgeneratedusingarandomselecGonofaXributesateachnodetodeterminethesplit
n DuringclassificaGon,eachtreevotesandthemostpopularclassisreturned
n TwoMethodstoconstructRandomForest:n Forest-RI(randominputselec4on):Randomlyselect,ateachnode,F
aXributesascandidatesforthesplitatthenode.TheCARTmethodologyisusedtogrowthetreestomaximumsize
n Forest-RC(randomlinearcombina4ons):CreatesnewaXributes(orfeatures)thatarealinearcombinaGonoftheexisGngaXributes(reducesthecorrelaGonbetweenindividualclassifiers)
n ComparableinaccuracytoAdaboost,butmorerobusttoerrorsandoutliersn InsensiGvetothenumberofaXributesselectedforconsideraGonateach
split,andfasterthanbaggingorboosGng67
![Page 68: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/68.jpg)
Classification of Class-Imbalanced Data Sets
n Class-imbalanceproblem:RareposiGveexamplebutnumerousnegaGveones,e.g.,medicaldiagnosis,fraud,oil-spill,fault,etc.
n TradiGonalmethodsassumeabalanceddistribuGonofclassesandequalerrorcosts:notsuitableforclass-imbalanceddata
n Typicalmethodsforimbalancedatain2-classclassificaGon:n Oversampling:re-samplingofdatafromposiGveclassn Under-sampling:randomlyeliminatetuplesfromnegaGveclass
n Threshold-moving:movesthedecisionthreshold,t,sothattherareclasstuplesareeasiertoclassify,andhence,lesschanceofcostlyfalsenegaGveerrors
n Ensembletechniques:EnsemblemulGpleclassifiersintroducedabove
n SGlldifficultforclassimbalanceproblemonmulGclasstasks68
![Page 69: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/69.jpg)
69
Chapter 8. Classification: Basic Concepts
n ClassificaGon:BasicConcepts
n DecisionTreeInducGon
n BayesClassificaGonMethods
n Rule-BasedClassificaGon
n ModelEvaluaGonandSelecGon
n TechniquestoImproveClassificaGonAccuracy:EnsembleMethods
n Summary
![Page 70: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/70.jpg)
Summary (I)
n ClassificaGonisaformofdataanalysisthatextractsmodelsdescribingimportantdataclasses.
n EffecGveandscalablemethodshavebeendevelopedfordecisiontreeinducGon,NaiveBayesianclassificaGon,rule-basedclassificaGon,andmanyotherclassificaGonmethods.
n EvaluaGonmetricsinclude:accuracy,sensiGvity,specificity,precision,recall,Fmeasure,andFßmeasure.
n StraGfiedk-foldcross-validaGonisrecommendedforaccuracyesGmaGon.BaggingandboosGngcanbeusedtoincreaseoverallaccuracybylearningandcombiningaseriesofindividualmodels.
70
![Page 71: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/71.jpg)
Summary (II)
n SignificancetestsandROCcurvesareusefulformodelselecGon.
n TherehavebeennumerouscomparisonsofthedifferentclassificaGonmethods;themaXerremainsaresearchtopic
n Nosinglemethodhasbeenfoundtobesuperioroverallothersforalldatasets
n Issuessuchasaccuracy,trainingGme,robustness,scalability,andinterpretabilitymustbeconsideredandcaninvolvetrade-
offs,furthercomplicaGngthequestforanoverallsuperiormethod
71
![Page 72: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/72.jpg)
References (1) n C.ApteandS.Weiss.Dataminingwithdecisiontreesanddecisionrules.Future
GeneraGonComputerSystems,13,1997n C.M.Bishop,NeuralNetworksforPa9ernRecogni/on.OxfordUniversityPress,
1995n L.Breiman,J.Friedman,R.Olshen,andC.Stone.Classifica/onandRegressionTrees.
WadsworthInternaGonalGroup,1984n C.J.C.Burges.ATutorialonSupportVectorMachinesforPa9ernRecogni/on.Data
MiningandKnowledgeDiscovery,2(2):121-168,1998n P.K.ChanandS.J.Stolfo.Learningarbiterandcombinertreesfrompar//oneddata
forscalingmachinelearning.KDD'95n H.Cheng,X.Yan,J.Han,andC.-W.Hsu,
Discrimina/veFrequentPa9ernAnalysisforEffec/veClassifica/on,ICDE'07n H.Cheng,X.Yan,J.Han,andP.S.Yu,
DirectDiscrimina/vePa9ernMiningforEffec/veClassifica/on,ICDE'08n W.Cohen.Fasteffec/veruleinduc/on.ICML'95n G.Cong,K.-L.Tan,A.K.H.Tung,andX.Xu.Miningtop-kcoveringrulegroupsfor
geneexpressiondata.SIGMOD'0572
![Page 73: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/73.jpg)
References (2) n A.J.Dobson.AnIntroduc/ontoGeneralizedLinearModels.Chapman&Hall,1990.n G.DongandJ.Li.Efficientminingofemergingpa9erns:Discoveringtrendsand
differences.KDD'99.n R.O.Duda,P.E.Hart,andD.G.Stork.Pa9ernClassifica/on,2ed.JohnWiley,2001n U.M.Fayyad.Branchingona9ributevaluesindecisiontreegenera/on.AAAI’94.n Y.FreundandR.E.Schapire.Adecision-theore/cgeneraliza/onofon-linelearningand
anapplica/ontoboos/ng.J.ComputerandSystemSciences,1997.n J.Gehrke,R.Ramakrishnan,andV.GanG.Rainforest:Aframeworkforfastdecisiontree
construc/onoflargedatasets.VLDB’98.n J.Gehrke,V.Gant,R.Ramakrishnan,andW.-Y.Loh,BOAT--Op/mis/cDecisionTree
Construc/on.SIGMOD'99.n T.HasGe,R.Tibshirani,andJ.Friedman.TheElementsofSta/s/calLearning:Data
Mining,Inference,andPredic/on.Springer-Verlag,2001.n D.Heckerman,D.Geiger,andD.M.Chickering.LearningBayesiannetworks:The
combina/onofknowledgeandsta/s/caldata.MachineLearning,1995.n W.Li,J.Han,andJ.Pei,CMAR:AccurateandEfficientClassifica/onBasedonMul/ple
Class-Associa/onRules,ICDM'01.73
![Page 74: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/74.jpg)
References (3)
n T.-S.Lim,W.-Y.Loh,andY.-S.Shih.Acomparisonofpredic/onaccuracy,complexity,andtraining/meofthirty-threeoldandnewclassifica/onalgorithms.MachineLearning,2000.
n J.Magidson.TheChaidapproachtosegmenta/onmodeling:Chi-squaredautoma/cinterac/ondetec/on.InR.P.Bagozzi,editor,AdvancedMethodsofMarkeGngResearch,BlackwellBusiness,1994.
n M.Mehta,R.Agrawal,andJ.Rissanen.SLIQ:Afastscalableclassifierfordatamining.EDBT'96.
n T.M.Mitchell.MachineLearning.McGrawHill,1997.n S.K.Murthy,Automa/cConstruc/onofDecisionTreesfromData:AMul/-
DisciplinarySurvey,DataMiningandKnowledgeDiscovery2(4):345-389,1998n J.R.Quinlan.Induc/onofdecisiontrees.MachineLearning,1:81-106,1986.n J.R.QuinlanandR.M.Cameron-Jones.FOIL:Amidtermreport.ECML’93.n J.R.Quinlan.C4.5:ProgramsforMachineLearning.MorganKaufmann,1993.n J.R.Quinlan.Bagging,boos/ng,andc4.5.AAAI'96.
74
![Page 75: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/75.jpg)
References (4) n R.RastogiandK.Shim.Public:Adecisiontreeclassifierthatintegratesbuildingand
pruning.VLDB’98.n J.Shafer,R.Agrawal,andM.Mehta.SPRINT:Ascalableparallelclassifierfordata
mining.VLDB’96.n J.W.ShavlikandT.G.DieXerich.ReadingsinMachineLearning.MorganKaufmann,
1990.n P.Tan,M.Steinbach,andV.Kumar.Introduc/ontoDataMining.AddisonWesley,
2005.n S.M.WeissandC.A.Kulikowski.ComputerSystemsthatLearn:Classifica/onand
Predic/onMethodsfromSta/s/cs,NeuralNets,MachineLearning,andExpertSystems.MorganKaufman,1991.
n S.M.WeissandN.Indurkhya.Predic/veDataMining.MorganKaufmann,1997.n I.H.WiXenandE.Frank.DataMining:Prac/calMachineLearningToolsand
Techniques,2ed.MorganKaufmann,2005.n X.YinandJ.Han.CPAR:Classifica/onbasedonpredic/veassocia/onrules.SDM'03n H.Yu,J.Yang,andJ.Han.ClassifyinglargedatasetsusingSVMwithhierarchical
clusters.KDD'03.
75
![Page 76: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/76.jpg)
![Page 77: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/77.jpg)
CS412 Midterm Exam Statistics
n OpinionQuesGonAnswering:n Likethestyle:70.83%,dislike:29.16%n Examishard:55.75%,easy:0.6%,justright:43.63%n Time:plenty:3.03%,enough:36.96%,not:60%n ScoredistribuGon:#ofstudents(Total:180)n >=90:24n 80-89:54n 70-79:46
n FinalgradingarebasedonoverallscoreaccumulaGonandrelaGveclassdistribuGons
77
n 60-69:37n 50-59:15n 40-49:2
n <40:2
![Page 78: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/78.jpg)
78
Issues: Evaluating Classification Methods
n Accuracyn classifieraccuracy:predicGngclasslabeln predictoraccuracy:guessingvalueofpredictedaXributes
n Speedn Gmetoconstructthemodel(trainingGme)n Gmetousethemodel(classificaGon/predicGonGme)
n Robustness:handlingnoiseandmissingvaluesn Scalability:efficiencyindisk-residentdatabasesn Interpretability
n understandingandinsightprovidedbythemodeln Othermeasures,e.g.,goodnessofrules,suchasdecisiontree
sizeorcompactnessofclassificaGonrules
![Page 79: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/79.jpg)
79
Predictor Error Measures
n Measurepredictoraccuracy:measurehowfaroffthepredictedvalueisfromtheactualknownvalue
n Lossfunc/on:measurestheerrorbetw.yiandthepredictedvalueyi’
n Absoluteerror:|yi–yi’|n Squarederror:(yi–yi’)2
n Testerror(generalizaGonerror):theaveragelossoverthetestsetn Meanabsoluteerror:Meansquarederror:
n RelaGveabsoluteerror:RelaGvesquarederror:
Themeansquared-errorexaggeratesthepresenceofoutliersPopularlyuse(square)rootmean-squareerror,similarly,rootrelaGve
squarederror
d
yyd
iii∑
=
−1
|'|
d
yyd
iii∑
=
−1
2)'(
∑
∑
=
=
−
−
d
ii
d
iii
yy
yy
1
1
||
|'|
∑
∑
=
=
−
−
d
ii
d
iii
yy
yy
1
2
1
2
)(
)'(
![Page 80: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/80.jpg)
80
Scalable Decision Tree Induction Methods
n SLIQ(EDBT’96—Mehtaetal.)n BuildsanindexforeachaXributeandonlyclasslistandthecurrentaXributelistresideinmemory
n SPRINT(VLDB’96—J.Shaferetal.)n ConstructsanaXributelistdatastructure
n PUBLIC(VLDB’98—Rastogi&Shim)n Integratestreespli^ngandtreepruning:stopgrowingthetreeearlier
n RainForest(VLDB’98—Gehrke,Ramakrishnan&GanG)n BuildsanAVC-list(aXribute,value,classlabel)
n BOAT(PODS’99—Gehrke,GanG,Ramakrishnan&Loh)n Usesbootstrappingtocreateseveralsmallsamples
![Page 81: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/81.jpg)
81
Data Cube-Based Decision-Tree Induction
n IntegraGonofgeneralizaGonwithdecision-treeinducGon(Kamberetal.’97)
n ClassificaGonatprimiGveconceptlevels
n E.g.,precisetemperature,humidity,outlook,etc.
n Low-levelconcepts,scaXeredclasses,bushyclassificaGon-trees
n SemanGcinterpretaGonproblems
n Cube-basedmulG-levelclassificaGon
n RelevanceanalysisatmulG-levels
n InformaGon-gainanalysiswithdimension+level
![Page 82: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/82.jpg)
82 82
Veri Ön İşleme (Data Preprocessing)
n Data Preprocessing: Giriş
n Veri Kalitesi
n Veri Ön işlemedeki ana işlemler
n Veri Temizleme (Data Cleaning)
n Veri Uyumu (Data Integration)
n Veri Küçültme (Data Reduction)
n Veri Dönüştürme ve Verinin Ayrıklaştırılması (Data
Transformation and Data Discretization)
![Page 83: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/83.jpg)
Data Warehouse: A Multi-Tiered Architecture
Data Warehouse
(Veri Ambarı)
Extract Transform Load Refresh
OLAP Engine
Analysis Query Reports Data mining
Monitor &
Integrator Metadata
Veri Kaynakları Front-End Tools
Serve
Data Marts
Operational DBs
Other sources
Data Storage
OLAP Server
![Page 84: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/84.jpg)
84
Veri Kalitesi ( Data Quality)
n Çok boyutlu olarak veri kalitesi kriterleri : Neden Ön işlem yapılır?
n Kesinlik (Accuracy) doğru ve yanlış veriler
n Tamamlık (Completeness) : kaydedilmemiş veya ulaşılamayan
veriler
n Tutarlılık (Consistency) verilerin bir kısmının güncel olmaması,
sallantıda veriler (dangling)
n Güncellik (Timeliness)
n İnandırıcılık (Believability)
n Yorumlanabilirlik (Interpretability): Verinin ne kadar kolay
anlaşılacağı
![Page 85: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/85.jpg)
85
Veri Ön İşleme İşlemleri
n Veri Temizleme (Data cleaning)
n Eksik verilerin doldurulması, gürültülü verilerin düzeltilmesi, aykırı verilerin (outlier) temizlenmesi, uyuşmazlıkların (inconsistencies) çözümlenmesi
n Veri Entegrasyonu (Data integration)
n Farklı veri kaynaklarının, Veri Küplerinin veya Dosyaların entegre olması
n Verinin Küçültülmesi (Data reduction)
n Boyut Küçültme (Dimensionality reduction)
n Sayısal Küçültme (Numerosity reduction)
n Verinin Sıkıştırılması (Data compression)
n Verinin Dönüştürülmesi ve Ayrıklaştırılması (Data transformation and data discretization)
n Normalleştirme (Normalization )
n Kavram Hiyerarşisi (Concept hierarchy generation)
![Page 86: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/86.jpg)
86 86
Veri Ön İşleme (Data Preprocessing)
n Data Preprocessing: Giriş
n Veri Kalitesi
n Veri Ön işlemedeki ana işlemler
n Veri Temizleme (Data Cleaning)
n Veri Uyumu (Data Integration)
n Veri Küçültme (Data Reduction)
n Veri Dönüştürme ve Verinin Ayrıklaştırılması (Data
Transformation and Data Discretization)
![Page 87: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/87.jpg)
87
Veri Temizleme (Data Cleaning)
n Gerçek hayattaki veriler kirlidir: Çok sayıda makine, insan veya bilgisayar hataları, iletim bozulmaları yaşanabilir.
n Eksik Veri (incomplete) bazı özelliklerin eksik olması (missing data), sadece birleşik verinin (aggregate) bulunması
n örn., Meslek=“ ” (girilmemiş) n Gülrültülü Veri (noisy): Gürültü, hata veya aykırı veriler bulunması
n örn., Maaş=“−10” (hata) n Tutarsız Veri (inconsistent): farklı kaynaklardan farklı veriler
gelmesi n Yaş=“42”, Doğum Tarihi=“03/07/2010” n Eski notlama “1, 2, 3”, yeni notlama “A, B, C” n Tekrarlı kayıtlarda uyuşmazlık
n Kasıtlı Problemler (Intentional) n Doğum tarihi bilinmeyen herkese 1 Ocak yazılması
![Page 88: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/88.jpg)
88
Eksik Veriler (Incomplete (Missing) Data)
n Veriye her zaman erişilmesi mümkün değildir n Örn., bazı kayıtların alın(a)mamış olması. Satış
sırasında müşterilerin gelir düzeyinin yazılmamış olması.
n Eksik veriler genelde aşağıdaki durumlarda olur: n Donanımsal bozukluklardan
n Uyuşmazlık yüzünden silinen veriler n Anlaşılamayan verilerin girilmemiş olması
n Veri girişi sırasında veriye önem verilmemiş olması n Verideki değişikliklerin kaydedilmemiş olması
n Eksik verilerin çözülmesi gerekir
![Page 89: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/89.jpg)
89
Eksik veriler nasıl çözülür?
n İhmal etme: Eksik veriler işleme alınmaz, yokmuş gibi davranılır. Kullanılan VM yöntemine göre sonuca etkileri bilinmelidir.
n Eksik verilerin elle doldurulması: her zaman mümkün değildir ve bazan çok uzun ve maliyetli olabilir
n Otomatik olarak doldurulması
n Bütün eksik veriler için yeni bir sınıf oluşturulması (“bilinmiyor” gibi)
n Ortalamanın yazılması
n Sınıf bazında ortalamaların yazılması
n Bayesian formül ve karar ağacı uygulaması
![Page 90: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/90.jpg)
90
Gürültülü Veri (Noisy Data)
n Gürültü (Noise): ölçümdeki rasgele oluşan değerler n Yanlış özellik değerleri aşağıdaki durumlarda oluşabilir:
n Veri toplama araçlarındaki hatalar n Veri giriş problemleri n Veri iletim problemleri n Teknoloji sınırları n İsimlendirmedeki tutarsızlıklar
n Veri temizlemesini gerektiren diğer durumlar n Tekrarlı kayıtlar n Eksik veriler n Tutarsız veriler
![Page 91: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/91.jpg)
91
Gürültülü Veri Nasıl Çözülür?
n Paketleme (Binning) n Veri sıralanır ve eşit frekanslarda paketlere bölünür. n Eksik veriler farklı yöntemlerle doldurulur:
n Mean n Median n Boundary
n Regrezisyon (Regression) n Regrezisyon fonksiyonlarına tabi tutularak eksik
verilerin girilmesi n Bölütleme (Kümeleme , Clustering)
n Aykırı verilerin bulunması ve temizlenmesi n Bilgisayar ve insan bilgisinin ortaklaşa kullanılması
n detect suspicious values and check by human (e.g., deal with possible outliers)
![Page 92: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/92.jpg)
92
Veri Temizleme Süreci n Verideki farklılıkların yakalanması
n Üst verinin (metadata) kullanılması (örn., veri alanı (domain, range) , bağlılık (dependency), dağılım (distribution)
n Aşırı yüklü alanlar (Field Overloading) n Veri üzerinde kural kontrolleri (unique, consecutive, null) n Ticari yazılımların kullanılması
n Bilgi Ovalaması (Data scrubbing): Basit alan bilgileri kurallarla kontrol etmek (e.g., postal code, spell-check)
n Veri Denetimi (Data auditing): veriler üzerinden kural çıkarımı ve kurallara uymayanların bulunması (örn., correlation veya clustering ile aykırıların (outliers) bulunması)
n Veri Göçü ve Entegrasyonu (Data migration and integration) n Data migration Araçları: Verinin dönüştürülmesine izin verir n ETL (Extraction/Transformation/Loading) Araçları: Genelde grafik
arayüzü ile dönüşümü yönetme imkanı verir n İki farklı işin entegre yürütülmesi
n Iterative / interactive (Örn.., Potter’s Wheels)
![Page 93: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/93.jpg)
Aşırı Yüklü Alanlar (Overloaded Fields)
n Aşırı Yüklü Alanların Temizlenmesi n Zincirleme (Chaining)
n Birleştirme (Coupling)
n Çok Amaçlılık (Multipurpose)
93
![Page 94: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/94.jpg)
Örnekler
94
Vijayshankar Raman and Joseph M. Hellerstein , Potter’s Wheel: An Interactive Data Cleaning System
berkeley
![Page 95: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/95.jpg)
95 95
Chapter 3: Data Preprocessing
n Data Preprocessing: An Overview
n Data Quality
n Major Tasks in Data Preprocessing
n Data Cleaning
n Data Integration
n Data Reduction
n Data Transformation and Data Discretization
n Summary
![Page 96: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/96.jpg)
96 96
Data Integration
n Data integration:
n Combines data from multiple sources into a coherent store
n Schema integration: e.g., A.cust-id ≡ B.cust-#
n Integrate metadata from different sources
n Entity identification problem:
n Identify real world entities from multiple data sources, e.g., Bill Clinton = William Clinton
n Detecting and resolving data value conflicts
n For the same real world entity, attribute values from different sources are different
n Possible reasons: different representations, different scales, e.g., metric vs. British units
![Page 97: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/97.jpg)
97 97
Handling Redundancy in Data Integration
n Redundant data occur often when integration of multiple databases
n Object identification: The same attribute or object may have different names in different databases
n Derivable data: One attribute may be a “derived” attribute in another table, e.g., annual revenue
n Redundant attributes may be able to be detected by correlation analysis and covariance analysis
n Careful integration of the data from multiple sources may help reduce/avoid redundancies and inconsistencies and improve mining speed and quality
![Page 98: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/98.jpg)
98
Correlation Analysis (Nominal Data)
n Χ2 (chi-square) test
n The larger the Χ2 value, the more likely the variables are related
n The cells that contribute the most to the Χ2 value are those whose actual count is very different from the expected count
n Correlation does not imply causality n # of hospitals and # of car-theft in a city are correlated
n Both are causally linked to the third variable: population
∑−
=Expected
ExpectedObserved 22 )(
χ
![Page 99: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/99.jpg)
99
Chi-Square Calculation: An Example
n Χ2 (chi-square) calculation (numbers in parenthesis are expected counts calculated based on the data distribution in the two categories)
n It shows that like_science_fiction and play_chess are correlated in the group
93.507840
)8401000(360
)360200(210
)21050(90
)90250( 22222 =
−+
−+
−+
−=χ
Play chess Not play chess Sum (row)
Like science fiction 250(90) 200(360) 450
Not like science fiction 50(210) 1000(840) 1050
Sum(col.) 300 1200 1500
![Page 100: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/100.jpg)
100
Correlation Analysis (Numeric Data)
n Correlation coefficient (also called Pearson’s product moment coefficient)
where n is the number of tuples, and are the respective means of A and B, σA and σB are the respective standard deviation of A and B, and Σ(aibi) is the sum of the AB cross-product.
n If rA,B > 0, A and B are positively correlated (A’s values increase as B’s). The higher, the stronger correlation.
n rA,B = 0: independent; rAB < 0: negatively correlated
BA
n
i ii
BA
n
i iiBA n
BAnban
BbAar
σσσσ )1()(
)1())((
11, −
−=
−
−−=
∑∑ ==
A B
![Page 101: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/101.jpg)
101
Visually Evaluating Correlation
Scatter plots showing the similarity from –1 to 1.
![Page 102: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/102.jpg)
102
Correlation (viewed as linear relationship)
n Correlation measures the linear relationship between objects
n To compute correlation, we standardize data objects, A and B, and then take their dot product
)(/))((' AstdAmeanaa kk −=
)(/))((' BstdBmeanbb kk −=
''),( BABAncorrelatio •=
![Page 103: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/103.jpg)
103
Covariance (Numeric Data)
n Covariance is similar to correlation
where n is the number of tuples, and are the respective mean or expected values of A and B, σA and σB are the respective standard deviation of A and B.
n Positive covariance: If CovA,B > 0, then A and B both tend to be larger than their expected values.
n Negative covariance: If CovA,B < 0 then if A is larger than its expected value, B is likely to be smaller than its expected value.
n Independence: CovA,B = 0 but the converse is not true: n Some pairs of random variables may have a covariance of 0 but are not
independent. Only under some additional assumptions (e.g., the data follow multivariate normal distributions) does a covariance of 0 imply independence
A B
Correlation coefficient:
![Page 104: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/104.jpg)
Co-Variance: An Example
n It can be simplified in computation as
n Suppose two stocks A and B have the following values in one week:
(2, 5), (3, 8), (5, 10), (4, 11), (6, 14).
n Question: If the stocks are affected by the same industry trends, will
their prices rise or fall together?
n E(A) = (2 + 3 + 5 + 4 + 6)/ 5 = 20/5 = 4
n E(B) = (5 + 8 + 10 + 11 + 14) /5 = 48/5 = 9.6
n Cov(A,B) = (2×5+3×8+5×10+4×11+6×14)/5 − 4 × 9.6 = 4
n Thus, A and B rise together since Cov(A, B) > 0.
![Page 105: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/105.jpg)
105 105
Chapter 3: Data Preprocessing
n Data Preprocessing: An Overview
n Data Quality
n Major Tasks in Data Preprocessing
n Data Cleaning
n Data Integration
n Data Reduction
n Data Transformation and Data Discretization
n Summary
![Page 106: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/106.jpg)
106
Data Reduction Strategies n Data reduction: Obtain a reduced representation of the data set that
is much smaller in volume but yet produces the same (or almost the same) analytical results
n Why data reduction? — A database/data warehouse may store terabytes of data. Complex data analysis may take a very long time to run on the complete data set.
n Data reduction strategies n Dimensionality reduction, e.g., remove unimportant attributes
n Wavelet transforms n Principal Components Analysis (PCA) n Feature subset selection, feature creation
n Numerosity reduction (some simply call it: Data Reduction) n Regression and Log-Linear Models n Histograms, clustering, sampling n Data cube aggregation
n Data compression
![Page 107: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/107.jpg)
107
Data Reduction 1: Dimensionality Reduction
n Curse of dimensionality n When dimensionality increases, data becomes increasingly sparse n Density and distance between points, which is critical to clustering, outlier
analysis, becomes less meaningful
n The possible combinations of subspaces will grow exponentially n Dimensionality reduction
n Avoid the curse of dimensionality n Help eliminate irrelevant features and reduce noise
n Reduce time and space required in data mining n Allow easier visualization
n Dimensionality reduction techniques n Wavelet transforms n Principal Component Analysis n Supervised and nonlinear techniques (e.g., feature selection)
![Page 108: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/108.jpg)
108
Mapping Data to a New Space
Two Sine Waves Two Sine Waves + Noise Frequency
n Fourier transform n Wavelet transform
![Page 109: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/109.jpg)
109
What Is Wavelet Transform?
n Decomposes a signal into different frequency subbands
n Applicable to n-dimensional signals
n Data are transformed to preserve relative distance between objects at different levels of resolution
n Allow natural clusters to become more distinguishable
n Used for image compression
![Page 110: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/110.jpg)
110
Wavelet Transformation
n Discrete wavelet transform (DWT) for linear signal processing, multi-resolution analysis
n Compressed approximation: store only a small fraction of the strongest of the wavelet coefficients
n Similar to discrete Fourier transform (DFT), but better lossy compression, localized in space
n Method: n Length, L, must be an integer power of 2 (padding with 0’s, when
necessary) n Each transform has 2 functions: smoothing, difference n Applies to pairs of data, resulting in two set of data of length L/2
n Applies two functions recursively, until reaches the desired length
Haar2 Daubechie4
![Page 111: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/111.jpg)
111
Wavelet Decomposition
n Wavelets: A math tool for space-efficient hierarchical decomposition of functions
n S = [2, 2, 0, 2, 3, 5, 4, 4] can be transformed to S^ = [23/4, -11/4, 1/2, 0, 0, -1, -1, 0]
n Compression: many small detail coefficients can be replaced by 0’s, and only the significant coefficients are retained
![Page 112: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/112.jpg)
112
Haar Wavelet Coefficients Coefficient “Supports”
2 2 0 2 3 5 4 4
-1.25
2.75
0.5 0
0 -1 0 -1
+
- +
+
+ + +
+
+
- -
- - - -
+
- +
+ - + -
+ - + -
- ++-
-1 -1
0.5
0
2.75
-1.25
0
0 Original frequency distribution
Hierarchical decomposition structure (a.k.a. “error tree”)
![Page 113: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/113.jpg)
113
Why Wavelet Transform?
n Use hat-shape filters n Emphasize region where points cluster n Suppress weaker information in their boundaries
n Effective removal of outliers n Insensitive to noise, insensitive to input order
n Multi-resolution n Detect arbitrary shaped clusters at different scales
n Efficient n Complexity O(N)
n Only applicable to low dimensional data
![Page 114: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/114.jpg)
114
x2
x1
e
Principal Component Analysis (PCA)
n Find a projection that captures the largest amount of variation in data n The original data are projected onto a much smaller space, resulting
in dimensionality reduction. We find the eigenvectors of the covariance matrix, and these eigenvectors define the new space
![Page 115: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/115.jpg)
115
n Given N data vectors from n-dimensions, find k ≤ n orthogonal vectors (principal components) that can be best used to represent data
n Normalize input data: Each attribute falls within the same range
n Compute k orthonormal (unit) vectors, i.e., principal components
n Each input data (vector) is a linear combination of the k principal component vectors
n The principal components are sorted in order of decreasing “significance” or strength
n Since the components are sorted, the size of the data can be reduced by eliminating the weak components, i.e., those with low variance (i.e., using the strongest principal components, it is possible to reconstruct a good approximation of the original data)
n Works for numeric data only
Principal Component Analysis (Steps)
![Page 116: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/116.jpg)
116
Attribute Subset Selection
n Another way to reduce dimensionality of data n Redundant attributes
n Duplicate much or all of the information contained in one or more other attributes
n E.g., purchase price of a product and the amount of sales tax paid
n Irrelevant attributes n Contain no information that is useful for the data
mining task at hand n E.g., students' ID is often irrelevant to the task of
predicting students' GPA
![Page 117: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/117.jpg)
117
Heuristic Search in Attribute Selection
n There are 2d possible attribute combinations of d attributes n Typical heuristic attribute selection methods:
n Best single attribute under the attribute independence assumption: choose by significance tests
n Best step-wise feature selection: n The best single-attribute is picked first n Then next best attribute condition to the first, ...
n Step-wise attribute elimination: n Repeatedly eliminate the worst attribute
n Best combined attribute selection and elimination n Optimal branch and bound:
n Use attribute elimination and backtracking
![Page 118: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/118.jpg)
118
Attribute Creation (Feature Generation)
n Create new attributes (features) that can capture the important information in a data set more effectively than the original ones
n Three general methodologies n Attribute extraction
n Domain-specific n Mapping data to new space (see: data reduction)
n E.g., Fourier transformation, wavelet transformation, manifold approaches (not covered)
n Attribute construction n Combining features (see: discriminative frequent
patterns in Chapter 7) n Data discretization
![Page 119: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/119.jpg)
119
Data Reduction 2: Numerosity Reduction
n Reduce data volume by choosing alternative, smaller forms of data representation
n Parametric methods (e.g., regression) n Assume the data fits some model, estimate model
parameters, store only the parameters, and discard the data (except possible outliers)
n Ex.: Log-linear models—obtain value at a point in m-D space as the product on appropriate marginal subspaces
n Non-parametric methods n Do not assume models n Major families: histograms, clustering, sampling, …
![Page 120: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/120.jpg)
120
Parametric Data Reduction: Regression and Log-Linear Models
n Linear regression n Data modeled to fit a straight line n Often uses the least-square method to fit the line
n Multiple regression n Allows a response variable Y to be modeled as a
linear function of multidimensional feature vector n Log-linear model
n Approximates discrete multidimensional probability distributions
![Page 121: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/121.jpg)
121
Regression Analysis
n Regression analysis: A collective name for techniques for the modeling and analysis of numerical data consisting of values of a dependent variable (also called response variable or measurement) and of one or more independent variables (aka.
explanatory variables or predictors)
n The parameters are estimated so as to give a "best fit" of the data
n Most commonly the best fit is evaluated by using the least squares method, but other criteria have also been used
n Used for prediction (including forecasting of time-series data), inference, hypothesis testing, and modeling of causal relationships
y
x
y = x + 1
X1
Y1
Y1’
![Page 122: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/122.jpg)
122
n Linear regression: Y = w X + b
n Two regression coefficients, w and b, specify the line and are to be estimated by using the data at hand
n Using the least squares criterion to the known values of Y1, Y2, …, X1, X2, ….
n Multiple regression: Y = b0 + b1 X1 + b2 X2
n Many nonlinear functions can be transformed into the above
n Log-linear models:
n Approximate discrete multidimensional probability distributions
n Estimate the probability of each point (tuple) in a multi-dimensional space for a set of discretized attributes, based on a smaller subset of dimensional combinations
n Useful for dimensionality reduction and data smoothing
Regress Analysis and Log-Linear Models
![Page 123: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/123.jpg)
123
Histogram Analysis
n Divide data into buckets and store average (sum) for each bucket
n Partitioning rules:
n Equal-width: equal bucket range
n Equal-frequency (or equal-depth)
05
1015
2025
3035
40
10000
20000
30000
40000
50000
60000
70000
80000
90000
100000
![Page 124: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/124.jpg)
124
Clustering
n Partition data set into clusters based on similarity, and store cluster representation (e.g., centroid and diameter) only
n Can be very effective if data is clustered but not if data is “smeared”
n Can have hierarchical clustering and be stored in multi-dimensional index tree structures
n There are many choices of clustering definitions and clustering algorithms
n Cluster analysis will be studied in depth in Chapter 10
![Page 125: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/125.jpg)
125
Sampling
n Sampling: obtaining a small sample s to represent the whole data set N
n Allow a mining algorithm to run in complexity that is potentially sub-linear to the size of the data
n Key principle: Choose a representative subset of the data
n Simple random sampling may have very poor performance in the presence of skew
n Develop adaptive sampling methods, e.g., stratified sampling:
n Note: Sampling may not reduce database I/Os (page at a time)
![Page 126: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/126.jpg)
126
Types of Sampling
n Simple random sampling n There is an equal probability of selecting any particular
item n Sampling without replacement
n Once an object is selected, it is removed from the population
n Sampling with replacement n A selected object is not removed from the population
n Stratified sampling: n Partition the data set, and draw samples from each
partition (proportionally, i.e., approximately the same percentage of the data)
n Used in conjunction with skewed data
![Page 127: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/127.jpg)
127
Sampling: With or without Replacement
SRSWOR
(simple random
sample without
replacement)
SRSWR
Raw Data
![Page 128: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/128.jpg)
128
Sampling: Cluster or Stratified Sampling
Raw Data Cluster/Stratified Sample
![Page 129: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/129.jpg)
129
Data Cube Aggregation
n The lowest level of a data cube (base cuboid)
n The aggregated data for an individual entity of interest
n E.g., a customer in a phone calling data warehouse
n Multiple levels of aggregation in data cubes
n Further reduce the size of data to deal with
n Reference appropriate levels
n Use the smallest representation which is enough to solve the task
n Queries regarding aggregated information should be answered using data cube, when possible
![Page 130: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/130.jpg)
130
Data Reduction 3: Data Compression
n String compression n There are extensive theories and well-tuned algorithms n Typically lossless, but only limited manipulation is
possible without expansion n Audio/video compression
n Typically lossy compression, with progressive refinement n Sometimes small fragments of signal can be
reconstructed without reconstructing the whole n Time sequence is not audio
n Typically short and vary slowly with time n Dimensionality and numerosity reduction may also be
considered as forms of data compression
![Page 131: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/131.jpg)
131
Data Compression
Original Data Compressed Data
lossless
Original Data Approximated
![Page 132: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/132.jpg)
132
Chapter 3: Data Preprocessing
n Data Preprocessing: An Overview
n Data Quality
n Major Tasks in Data Preprocessing
n Data Cleaning
n Data Integration
n Data Reduction
n Data Transformation and Data Discretization
n Summary
![Page 133: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/133.jpg)
133
Data Transformation n A function that maps the entire set of values of a given attribute to a
new set of replacement values s.t. each old value can be identified with one of the new values
n Methods
n Smoothing: Remove noise from data
n Attribute/feature construction
n New attributes constructed from the given ones
n Aggregation: Summarization, data cube construction
n Normalization: Scaled to fall within a smaller, specified range
n min-max normalization
n z-score normalization
n normalization by decimal scaling
n Discretization: Concept hierarchy climbing
![Page 134: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/134.jpg)
134
Normalization n Min-max normalization: to [new_minA, new_maxA]
n Ex. Let income range $12,000 to $98,000 normalized to [0.0, 1.0]. Then $73,000 is mapped to
n Z-score normalization (µ: mean, σ: standard deviation):
n Ex. Let µ = 54,000, σ = 16,000. Then
n Normalization by decimal scaling
716.00)00.1(000,12000,98000,12600,73
=+−−
−
AAA
AA
A minnewminnewmaxnewminmaxminvv _)__(' +−−
−=
A
Avvσµ−
='
j
vv10
'= Where j is the smallest integer such that Max(|ν’|) < 1
225.1000,16
000,54600,73=
−
![Page 135: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/135.jpg)
135
Discretization n Three types of attributes
n Nominal—values from an unordered set, e.g., color, profession
n Ordinal—values from an ordered set, e.g., military or academic rank
n Numeric—real numbers, e.g., integer or real numbers
n Discretization: Divide the range of a continuous attribute into intervals
n Interval labels can then be used to replace actual data values
n Reduce data size by discretization
n Supervised vs. unsupervised
n Split (top-down) vs. merge (bottom-up)
n Discretization can be performed recursively on an attribute
n Prepare for further analysis, e.g., classification
![Page 136: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/136.jpg)
136
Data Discretization Methods
n Typical methods: All the methods can be applied recursively
n Binning
n Top-down split, unsupervised
n Histogram analysis
n Top-down split, unsupervised
n Clustering analysis (unsupervised, top-down split or bottom-up merge)
n Decision-tree analysis (supervised, top-down split)
n Correlation (e.g., χ2) analysis (unsupervised, bottom-up merge)
![Page 137: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/137.jpg)
137
Simple Discretization: Binning
n Equal-width (distance) partitioning
n Divides the range into N intervals of equal size: uniform grid
n if A and B are the lowest and highest values of the attribute, the
width of intervals will be: W = (B –A)/N.
n The most straightforward, but outliers may dominate presentation
n Skewed data is not handled well
n Equal-depth (frequency) partitioning
n Divides the range into N intervals, each containing approximately
same number of samples
n Good data scaling
n Managing categorical attributes can be tricky
![Page 138: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/138.jpg)
138
Binning Methods for Data Smoothing
q Sorted data for price (in dollars): 4, 8, 9, 15, 21, 21, 24, 25, 26, 28, 29, 34
* Partition into equal-frequency (equi-depth) bins: - Bin 1: 4, 8, 9, 15 - Bin 2: 21, 21, 24, 25 - Bin 3: 26, 28, 29, 34 * Smoothing by bin means: - Bin 1: 9, 9, 9, 9 - Bin 2: 23, 23, 23, 23 - Bin 3: 29, 29, 29, 29 * Smoothing by bin boundaries: - Bin 1: 4, 4, 4, 15 - Bin 2: 21, 21, 25, 25 - Bin 3: 26, 26, 26, 34
![Page 139: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/139.jpg)
139
Discretization Without Using Class Labels
(Binning vs. Clustering)
Data Equal interval width (binning)
Equal frequency (binning) K-means clustering leads to better results
![Page 140: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/140.jpg)
140
Discretization by Classification & Correlation Analysis
n Classification (e.g., decision tree analysis)
n Supervised: Given class labels, e.g., cancerous vs. benign
n Using entropy to determine split point (discretization point)
n Top-down, recursive split
n Details to be covered in Chapter 7
n Correlation analysis (e.g., Chi-merge: χ2-based discretization)
n Supervised: use class information
n Bottom-up merge: find the best neighboring intervals (those
having similar distributions of classes, i.e., low χ2 values) to merge
n Merge performed recursively, until a predefined stopping condition
![Page 141: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/141.jpg)
141
Concept Hierarchy Generation
n Concept hierarchy organizes concepts (i.e., attribute values) hierarchically and is usually associated with each dimension in a data warehouse
n Concept hierarchies facilitate drilling and rolling in data warehouses to view data in multiple granularity
n Concept hierarchy formation: Recursively reduce the data by collecting and replacing low level concepts (such as numeric values for age) by higher level concepts (such as youth, adult, or senior)
n Concept hierarchies can be explicitly specified by domain experts and/or data warehouse designers
n Concept hierarchy can be automatically formed for both numeric and nominal data. For numeric data, use discretization methods shown.
![Page 142: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/142.jpg)
142
Concept Hierarchy Generation for Nominal Data
n Specification of a partial/total ordering of attributes explicitly at the schema level by users or experts
n street < city < state < country n Specification of a hierarchy for a set of values by explicit
data grouping n {Urbana, Champaign, Chicago} < Illinois
n Specification of only a partial set of attributes n E.g., only street < city, not others
n Automatic generation of hierarchies (or attribute levels) by the analysis of the number of distinct values
n E.g., for a set of attributes: {street, city, state, country}
![Page 143: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/143.jpg)
143
Automatic Concept Hierarchy Generation
n SomehierarchiescanbeautomaGcallygeneratedbasedontheanalysisofthenumberofdisGnctvaluesperaXributeinthedatasetn TheaXributewiththemostdisGnctvaluesisplacedatthelowestlevelofthehierarchy
n ExcepGons,e.g.,weekday,month,quarter,year
country
province_or_ state
city
street
15 distinct values
365 distinct values
3567 distinct values
674,339 distinct values
![Page 144: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/144.jpg)
144
Chapter 3: Data Preprocessing
n Data Preprocessing: An Overview
n Data Quality
n Major Tasks in Data Preprocessing
n Data Cleaning
n Data Integration
n Data Reduction
n Data Transformation and Data Discretization
n Summary
![Page 145: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/145.jpg)
145
Summary n Data quality: accuracy, completeness, consistency, timeliness,
believability, interpretability n Data cleaning: e.g. missing/noisy values, outliers n Data integration from multiple sources:
n Entity identification problem n Remove redundancies n Detect inconsistencies
n Data reduction n Dimensionality reduction n Numerosity reduction n Data compression
n Data transformation and data discretization n Normalization n Concept hierarchy generation
![Page 146: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/146.jpg)
Kaynaklar
n Data Mining: Concepts and Techniques, Third Edition (The Morgan Kaufmann Series in Data Management Systems) 3rd Edition
n by Jiawei Han (Author), Micheline Kamber (Author), Jian Pei (Author)
![Page 147: iticu dm 4 classification youtube - Dr. Sadi Evren SEKER...5 n Classificaon n predicts categorical class labels (discrete or nominal) n classifies data (constructs a model) based](https://reader033.fdocuments.us/reader033/viewer/2022053122/60a80a54e8d966580e2c9b82/html5/thumbnails/147.jpg)
147
References n D.P.BallouandG.K.Tayi.Enhancingdataqualityindatawarehouseenvironments.Comm.of
ACM,42:73-78,1999n A.Bruce,D.Donoho,andH.-Y.Gao.Waveletanalysis.IEEESpectrum,Oct1996n T.DasuandT.Johnson.ExploratoryDataMiningandDataCleaning.JohnWiley,2003n J.DevoreandR.Peck.Sta4s4cs:TheExplora4onandAnalysisofData.DuxburyPress,1997.n H.Galhardas,D.Florescu,D.Shasha,E.Simon,andC.-A.Saita.DeclaraGvedatacleaning:
Language,model,andalgorithms.VLDB'01n M.HuaandJ.Pei.Cleaningdisguisedmissingdata:AheurisGcapproach.KDD'07n H.V.Jagadish,etal.,SpecialIssueonDataReducGonTechniques.BulleGnoftheTechnical
CommiXeeonDataEngineering,20(4),Dec.1997n H.LiuandH.Motoda(eds.).FeatureExtrac4on,Construc4on,andSelec4on:ADataMining
Perspec4ve.KluwerAcademic,1998n J.E.Olson.DataQuality:TheAccuracyDimension.MorganKaufmann,2003n D.Pyle.DataPreparaGonforDataMining.MorganKaufmann,1999n V.RamanandJ.Hellerstein.PoXersWheel:AnInteracGveFrameworkforDataCleaningand
TransformaGon,VLDB’2001n T.Redman.DataQuality:TheFieldGuide.DigitalPress(Elsevier),2001n R.Wang,V.Storey,andC.Firth.Aframeworkforanalysisofdataqualityresearch.IEEETrans.
KnowledgeandDataEngineering,7:623-640,1995