CharlesBerger

Directorof

ProductManagementDataMiningTechnologiesOracleCorporationOracleDataMiningSolutionsAgendaDataMiningOverviewOracleDataMiningSuite(OracleDarwin)OracleDataMiningSuiteRelease3.7CRM11iintegrationOraclePersonalizationFutureDirections:OracleDataMinerDataMiningOverviewWhatIsDataMining?“Simplyput,dataminingisusedto

discover[hidden]patterns

andrelationships

inyourdatainordertohelpyoumake

betterbusinessdecisions.”

--

RobertSmall,TwoCrowsEnablingPersonalizedRelationships1:1RelationshipsUnderstandand

predictindividual

customerbehaviorOfferproductsandservicesthatanticipatecustomerneedsBuildloyaltyand

increaseprofitabilityWantsanewcellphoneHastwodaughtersTravelsacrossstatelinesfrequently

Offerher:1.Wide-areadigitalphoneplan2.Emergency-useplanfordaughtersAdHocQuery&Reporting,OLAP,andDataMiningDataMining

Insight:Highlyaccuratepredictionsofcustomerbehavior,wherestatisticaltechniquesfallshortForesight

(OLAP)ActualCostForecast10,000,0007,500,000TOTALS$73,996,564.06

JUL98SEP98NOV98ActualCostForecastExpenses10,000,0007,500,000Foresight

(OLAP)PercentChurnedAllPromotionPlans49.8038.3940.9240.1340.13DecisionTree-SegmentationIncomeMonthsasCustomerCHURNERSvs.

LOYALCUSTOMERSInsight

(Datamining)Hindsight(Ad-hocquery)DataMiningConceptsExample“hiddenpattern”foundbydataminingWhoislikely

tobuya

Buick?Parents

bought“foreign”EntirePopulationAge>45FemaleAge<45“Yes”/.55ConfidenceMale“Yes”/.67ConfidenceParents

ownedBuick

“Yes”/.92ConfidenceDataMiningConceptsIndependentVariablesX1X2XmYDependent

VariableFunctional

Relationship:Y=F(X1,X2,…,Xm)AttributesorFieldsRecords(targetvalue(Y))NameIncomeAge.......BuyaBuick?1=Yes,0=NoJonesSmithLeeRogers30,00055,00025,00050,000306723441100ModelPredictions1010Confidence.85.74.93.65CorporateDataOracleDataMiningSuite(Darwin)

PowerandEaseofUseIntuitiveGUI“Wizards”toguide

andautomatePowerful,scalable,parallelUNIXserverIntegratedwithin

Oracle“stack”MSExcelintegrationClient:WindowsServers:

SunSolarisHP-UXOracleDataMiningEnvironmentOracleCustomerTableModelTableProfilesTableScoresTableModelSchemaDataMiningSuiteExpressDiscovererReportsApplicationsOracleDataMiningSuiteDifferentiatorsIntegratedwithOracleInternetPlatformEliminatesdatamovement

andredundancyEnterprise-wide,parallel, scalabledatamining

solutionRapidlyextractsbusiness

intelligencehiddenwithin

customerdataComprehensive,

multi-algorithmicapproachMinesmoredatatobetterunderstandandpredictcustomerbehaviorEaseofUseIntuitiveGUIIntegrateswithexistingsystemstodistribute

businessintelligencethroughouttheenterpriseCumulativeTargetsvs.PercentofPopulationCumulativeTargetsMostPopularIndustriesBusiness-to-Consumer-orientedcompanieswithlargeamountsofdataHundredsoffields(elements)HundredsofmillionsofrecordsIndustriesTelcosBanks/FinancialServicesInsurance/HealthcareGovernmentDatabaseMarketingDot.Com&e-BusinessOtherindustries:Airlines,Hotels,MfgPopularDataMiningApplicationsProfilingChurn/AttritionRetentionCross-SellingResponseModeling/TargetMarketingProfitabilityAnalysisProductAffinityBundlesFraudDetectionCallCenters/TelemarketingOracleDataMining3.7NewFeatures

AvailableOctober,2000OCIconnectivityforfasterdataloadingandexportingSupporttomine

8-bitWesternEuropeandatacharactersetsCRM11iIntegrationCRM11iIntegrationCRM11iIntegrationCRM11iIntegrationAdditionalIntegrationPossibilitiesClickstreamAnalyzeriStoreiASWirelessSalesForceAutomationOracleBusinessIntelligencemodulesCustomerIntelligenceMarketingIntelligenceHRIntelligenceQualityIntelligenceetc.Announcing

OraclePersonalization

Real-TimeRecommendationOfferinginOracle9i

“Brickandmortar”storeshave

anadvantage:thestoreclerkSuggestappropriateproductsbased

uponthecustomer’sstatedpreferencesDataminingtotherescueThe“e-salesclerk”metaphorThe“e-salesclerk”can:Trackcustomers’movementRecordcustomers’preferencesApplyminingmodelstorecommend

products,links,andbetternavigationThe“e-salesclerk”emulatesa

knowledgeablestoreclerkWhoIsMindingtheVirtualE-Store?ThePromiseofWebMiningHello,MarkJones!Wehave

recommendationsBooksR’sBookStore-NetscapeBooksRus.comBooksRAuctions.BooksR100HotBooksHello,MarkJones.Wethinkyouwillliketheseitems.Iam

lookingforadataminingbookRecommendationforadditionalbooksthatmatchhisinterestsandprofilePersonalizationUsingWebMiningClickstream—WebographicsInquiries/SearchRegistrationInformationCorporateDBs—Demographics$$$WebIntegrateContentNavigationAdsServiceSearchresultsCross-sellingLinksShortcutsProductsPromotionsMailingsCallcenterCustomerCustomize&PersonalizeOraclePersonalization

Oracle9iServerFeaturesMiningperformedwithindatabaseAlgorithmsTransactionalNa?veBayesPredictiveAssociationRulesPL/SQLmodelsDatabase-internalscoringRecommendationTasksWhichNitemsispersonAmostlikelytobuyorlike?PeoplethatboughtorlikeitemXarelikelytobuyorlikewhichotheritem(s)?HowlikelyispersonAtobuyorlikeitemX?WhichNitemsispersonAmostlikelytobuyorlike,givenhelikesorisbuyinganotheritem?WhichNitemsaremostsimilartoitemXalongagivendimension?ModelingStrategiesRecommendationstasksaresupportedbybatchandonlinescoringusinginductivemodelsPeoplethatboughtorlikeitemXarelikelytobuyorlikewhichotheritem(s)?Databasequeries—pre-scoredresultsstoredinthedatabaseWhichNitemsispersonAmostlikelytobuyorlike?Inductivemodeling—onlinescoringusingcustomerprofiles(historicalandcurrentdata)ArchitectureHistoricaldataWebApplicationMobileApplicationCallCenterApplicationCampaignManagementRequestsfor

Recommendations

OraclePersonalizationOracle9iPredictive

ModelsRecommendationEngineFarmsRecommendationsMovieT:ExampleDemoNewWorldCoffee:ExampleDemoVisitorlogsintoWebsiteNewWorldCoffee:ExampleDemoVisitorclickson“Products”NewWorldCoffee:ExampleDemoVisitorclickson“Coffee”NewWorldCoffee:ExampleDemoVisitorclickson“Mints”NewWorldCoffee:ExampleDemoBanneradsmakereal-time“personal”recommendationsMintsthatarelikelytobefavoredbythisvisitorarerecommendedOraclePersonalization

MainActivitiesandRelease1AssumptionsOraclePersonalizationsetupbyDBAandWebmasterBuildDeploymentReportsTuningRelease1SetupactivityisimplementedatinstallationtimewithconsultingservicesManagementof“OraclePersonalizationSchema”

(taxonomy,keywords,dimensions,datatypes)isdonethroughDBAmanagementoftherelevanttablesOraclePersonalization

UserInterfaceOverview

HomePageCreateaDeploymentPackageSpecifyBuildandTuningSettingsSchedulePackageBuildSchedulePackageDeploymentViewReportsFutureDirections:

OracleDataMiner

MovingAdditionalDataMiningAlgorithmsintoOracle9i

Darwin3.X–AnalyticalToolExposesallthebellsandwhistlesFulladhocaccesstoatomicoperationsUsercanoverridealltooldecisionsEmphasisondataminingfeaturesFullsupportofdataminingprocessSupportformultipleskilllevels

(fromwizardstofullcontrol)OracleDataMiner

BusinessObjectivesBusinesstool—methodology-drivenSolvebusinessproblemsHidethedataminingdetailsToolmakesthetechnicaldecisionsDatabasemarketer/analystmakesthebusinessdecisions(versusadedicateddataminer)FocusondeploymentandapplicationintegrationOracleDataMiner

DesignObjectivesDatabase-focusedMiningandscoringperformedwithindatabaseBusinessproblem-oriented—nottool-orientedWeb-basedGUI,thinclienttechnologyJavaclientAPIandXMLforscriptingTightintegrationwithotherOracleproductsTranslatabletolocallanguagesOracleDataMiner

TargetCY01Thinclienttargetedfordatabasemarketer“Methodologies”orientation:GeneraldataminingRetention/churnResponsemodelingCustomerprofilingJavaAPIforApplications

(e.g.CRM)DevelopersOracleDataMiner

TargetCY01OracleDataMiner

TargetCY01MigrationtotheDatabaseandtheWebDarwin3.XServerDarwin3.XGUIBrowserIn-DBMiningO9iO8iEnablinge-BusinessIntelligence

MiningtheIntelligentWebHouseOracle9iCWMandRepositoryAnalyticApplicationsDiscovererReportsExpressServerEnterpriseManagerandWorkflowExpressToolsERP

DataOperationalDataExternalDataWeb

DataOracle