DMDW Lesson 05 + 06 + 07 - Data Mining Applied

STUDIERENUND DURCHSTARTEN.Author I: Dip.-Inf. (FH) Johannes HoppeAuthor II: M.Sc. Johannes HofmeisterAuthor III: Prof. Dr. Dieter HomeisterDate: 01.04.201108.04.201115.04.2011

Data Mining AppliedAuthor I: Dip.-Inf. (FH) Johannes HoppeAuthor II: M.Sc. Johannes HofmeisterAuthor III: Prof. Dr. Dieter HomeisterDate: 01.04.201108.04.201115.04.2011

01Applications of Data Mining3

Applicationsof Data MiningApplications of Data MiningDatabase Marketing Time-series prediction, detecting "trends" Detection (of whatever is detectable)Probability Estimation Information compression Sensitivity Analysis 5

Applicationsof Data MiningDatabase Marketing(1/2)Response modelingModel for the response of specific customers. Systematic selection of (old and potential) customers. Advertisements and promotion based on these results. ( CRM)Visualization: "Lift chart" shows how successful the selection should be. (later topic: DM validation)6

Lift Chart Example“For contacting 10% of customers, using no model we should get 10% of responders and using the given model we should get 30% of responders.”7

Applicationsof Data MiningDatabase Marketing(2/2)Cross selling: Selling additional products to existing customersQuestion: Which customer might buy which other product?Uses historical purchase data Uses credit card information, lifestyle data, demographic data, etc. Other possible information: Did the customer query special information? How customer heard of the company? 8

Applicationsof Data MiningDatabase Marketing(2/2)Cross selling: Selling additional products to existing customersResults for direct marketing, mailing lists, direct advertising (Amazon) Amazon: "Customers who bought this item also bought" and "personalized recommendations" 9

Applicationsof Data MiningTime-series predictionTime series: Stock prices, market shares, … Extrapolation of future values Detection of newly arising trends like customer movements to other productsOwn experience: German print magazines 10

Applicationsof Data MiningDetectionIdentification of existence or occurrence of a condition Fraud detection: Identifying patterns/criteria to detect credit card fraud Estimating creditworthiness ( German Schufa) Prediction of mail orders that will not be paid 11

Applicationsof Data MiningDetectionIdentification of existence or occurrence of a condition Intrusion detection (in computer networks) Find patterns that indicate when an attackis made on an network e.g. clustering: small clusters are of high interest,they point to unusual cases.Definition of Classes may be useful:e.g. harmless, possible harmful,harmful, immediately close LAN 12

Applicationsof Data MiningDetectionIdentification of existence or occurrence of a condition Typical difficultiesNeeds knowledgeDM costs Cost of missing a fraud Cost of false positives(e.g. falsely accusing someone of fraud, company image problems)13

Applicationsof Data MiningProbability EstimationApproximate the likelihood of an event given an observation e.g. for classify a potential customer into an A,B,C range before any business14

Applicationsof Data MiningInformation CompressionCan be viewed as a special type of estimation problem. For a given set of data, estimate the key components that be can be used to construct the data. 15

Applicationsof Data MiningSensitivity AnalysisUnderstand how changes in one variable affect others. Identify sensitivity of one variable on another(find out if dependencies exist). 16

Data Mining AlgorithmsData Mining AlgorithmsDifferent algorithms, different usesCombinedThe algorithm depends on what you want to doNot every algorithm is suited for what you want to do18

Data Mining AlgorithmsAlgorithms in SSAS: GroupsClassification algorithmsRegression algorithmsAssociation algorithmsSegmentation algorithmsSequence analysis algorithmsPlug-In algorithms19

Data Mining AlgorithmsClassification algorithmsPredict discrete attributesBased on experience valuesAlgorithms in SSAS:Naive BayesDecision TreesNeural Networks20

Data Mining AlgorithmsRegression algorithmsPredict continuous attributesThe same as classification algorithmsAlgorithms in SSASLinear Regression (Line)Logistic Regression (Curve)MS Time Series21

Data Mining AlgorithmsAssociation algorithmsPredict likely combinationsFind elements that occur in combinationAlgorithms in SSAS:MS Associtation Algorithm (Apriori)22

Data Mining AlgorithmsSegmentation algorithmsAlso called „Clustering algorithms“Groups data with similar propertiesAlgorithms in SSAS:MS Clustering Algorithms (e.g. K-Means)23

Data Mining AlgorithmsSequence analysis algorithms…are clustering algorithmsConsider the sorting; the sequence of values while clusteringDoes not group by similar propertiesGroups by similar sequencesAlgorithms in SSAS: MS Sequence Clustering24

Data Mining AlgorithmsPlug-In algorithms.NET Wrapper for COM objectsUse ANY algorithmProvided as an assembly(possible workshop to create one)25

03Repetition - Datatypes, Contentypes26

Repetition - Datatypes, ContentypesApplying anAlgorithmDatatypesContenttypes27

Repetition - Datatypes, ContentypesDatatypesDefinethestructure of thevaluesAvailabledatatypes:TextLongBooleanDoubleDate28

Repetition - Datatypes, ContentypesContenttypesDefinethebehaviour of valuesDiscreteContinuousDiscretizedKeyKey SequenceKey TimeOrderedCyclical29

Repetition - Datatypes, ContentypesContenttype: DiscreteFixed set of valuesExample:Commute Distance: 1-2, 2-5, 5-10Region: Pacific, Northern America, EuropeName: … … …Boolean values are always discreteText is most likely discrete30

Repetition - Datatypes, ContentypesContenttype: ContinuousUnlimited set of valuesInfinite items possibleExampleIncomeAgeDifference between Continuous and Discrete is the most important one31

Repetition - Datatypes, ContentypesContenttype: DiscretizedContinuousvaluesconvertedintodiscretevaluesExamples:Income to Categories:A, B, C, …Age to groups:0-20,21-30, 31-40, …32

Repetition - Datatypes, ContentypesContenttype: KeyKeyUniquely identifies a rowKey Sequence (sequence clustering models)Series of eventsSortedKey Time (time series models)Identify values on a time scale33

Repetition - Datatypes, ContentypesContenttype: OrderedDiscretevaluesthathave a sorting orderNodistancesvisibleNorelationsvisible„One Star“ to „Five Stars“34

Repetition - Datatypes, ContentypesContenttype: CyclicalDiscretevaluesthathave a cyclicalsorting orderExample:Weekdays: Monday, Tuesday, … Sunday, Monday, … 1,2,3, …,7, 1, …Months Jan, Feb, Mar, … , Dec, Jan, … 1, 2, 3, …, 12, 1, …35

04Data Mining Algorithms - Decision Trees37

Applied Data Mining - Decision Trees38

Applied Data Mining - Decision TreesIn GeneralAlso known as: Classification TreesGoal: Sequentially partition DataCan detect non-linear relationshipsMachine Learning TechniqueSeparate into Training and Testing setTraining set is created to create model based on certain criteriaTest set is used to verify the model39

Applied Data Mining - Decision TreesTree for response of a mailing actionIncome > $30 000: 3,6 %Male 3,2%(Total: 4.677)Income < $30 000: 2,3 %2,6 % respose rate(Total: 10.000 persons)Age > 40: 3,8%Female 2,1%(Total: 5,323)Age < 40: 3,2 %40

Applied Data Mining - Decision TreesUsingtheTrainedTreeExample: the management decides to mail only to groups with response rate >3.5%. TrainedTreeMales: $30 000Response Rate: > 3,5 %Female: 40+41

Applied Data Mining - Decision TreesProsVery flexible, white box ModelKiss – Keep it simple, stupid!Little preparation and resources neededConsCan be tuned until deathLong time to buildRequires wisely selected training data!False training yields false resultsBig tree might require disk swapping(Computation might be difficult if it does not fit into main memory.) 42

Project: “DMDW Mining Test”43

Project: “DMDW Mining Test”(explanation of one note)44

Project: “DMDW Mining Test”(shows connections, more useful if there are more predictable values)

Project: “DMDW Mining Test”(Generic Content Tree Viewer  DMX (Data Mining Extensions))

ReferencesReferences for Decisions TreesOlivia Parr Rud et. al, Data Mining Cookbook - Modeling Data for Marketing, Risk, and Customer Relationship Management, Wiley, 2001David A. Grossman, Ophir Frieder: Introductionto Data Mining, Illinois Institute of Technology 2005Andrew W. Moore: DecisionTrees, Carnegie Mellon University, http://www.autonlab.org/tutorials/dtree16.pdfNongYe (ed.): The Handbook of Data Mining, Lawrence Erlbaum Associates, 2003Sushimita Mitra, TinkuAcharya, Data Mining - Multimedia, Soft Computing andBioinformatics, Wiley, 2003http://en.wikipedia.org/wiki/Classification_tree47

05Data Mining Algorithms - Clustering48

Data Mining Algorithms - ClusteringX1249

Data Mining Algorithms - ClusteringClusteringSegmentation AlgorithmFind homogenous groups within setFind similar variables for different casesIdentify new relationships that were unclear before(heuristics)e.g. „Person who rides a bike to work doesn‘t live far from his workplace“ (this is not obvious)50

51HomogeneousSubsetsIndependent VariablesDescription of classclassifyidentifyX12

52HomogeneousSubsetsIndependent VariablesDescription of class1. Clustering2. ClassificationclassifyidentifyX12

Clustering1. ClusteringReducesdatatoclasses of equaltypesBecomefriedswiththedataIterative AlgorithmClusteringValidateClassifyApplyhttp://msdn.microsoft.com/en-us/library/ms174879.aspx53

Data Mining Algorithms - Clustering2. ClassificationCreate a Description of a groupGive it a „name“Also: Characterization54

ProcessStart with random valuesReuse will create different sets and different groupsDifferent clustering technique / algorithm will create different groupReuse on same dataset, reseedExpert evaluate found classes and plausibility Good classes used for predictionsGood?1. ClusteringEvaluate, Check2. ClassifyApply(Predict)55

ClusteringMS Clustering AlgorithmCombination of two algorithmsK-Means – Hard! Datapoint can be in only one clusterExpectation Maximization – SoftDatapoint has different combinationsDatapoint belongs to different clustersProbability is calculated56Source: http://msdn.microsoft.com/en-us/library/cc280445.aspx

Clustering57ProsNo predictable variable to chooseTrains itself without much effortEasy to configure„Cons“Interpretation is everythingGood eye neededExpert has to check for plausibility

Project: “DMDW Mining Test”(strongest relations only, amount of matching cases for Region Europe)

Project: “DMDW Mining Test”(good to know: continuous attributes are shown by there arithmetic average)

Project: “DMDW Mining Test”(comparing two clusters)

DMDW Lesson 05 + 06 + 07 - Data Mining Applied

More Related Content

What's hot (20)

Viewers also liked (8)

Similar to DMDW Lesson 05 + 06 + 07 - Data Mining Applied (20)

More from Johannes Hoppe (20)

Recently uploaded (20)

DMDW Lesson 05 + 06 + 07 - Data Mining Applied