数据挖掘概念与技术 第二版 韩家炜 第十一章a1visualmine

《数据挖掘概念与技术 第二版 韩家炜 第十一章a1visualmine》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、DataMining:ConceptsandTechniques—Chapter11——ApplicationsandTrendsinDataMining—AdditionalTheme:VisualDataMiningJiaweiHanandMichelineKamberDepartmentofComputerScienceUniversityofIllinoisatUrbana-Champaignwww.cs.uiuc.edu/~hanj©2006JiaweiHanandMichelineKamber.Allrigh

2、tsreserved.6/25/20211DataMining:ConceptsandTechniques6/25/20212DataMining:ConceptsandTechniquesVisualDataMining:AnOverviewWhatisVisualDataMining?SurveyoftechniquesDataVisualizationVisualizingDataMiningResultsVisualDataMining6/25/20213DataMining:ConceptsandTechniq

3、uesWhatIsVisualDataMining?Visualdatamining“discoversimplicitandusefulknowledgefromlargedatasetsusingdataand/orknowledgevisualizationtechniques”Datavisualization+Dataminingtechniques6/25/20214DataMining:ConceptsandTechniquesWhyVisualDataMining?Advantagesofhumanvis

4、ualsystemHighlyparallelprocessorSophisticatedreasoningengineLargeknowledgebaseCanbeusedtocomprehenddatadistributions,patterns,clusters,andoutliersDataMiningAlgorithmsVisualizationActionable+–Evaluation+–Flexibility–+UserInteraction–+6/25/20215DataMining:Conceptsa

5、ndTechniquesWhyNotOnlyVisualDataMining?DisadvantagesofhumanvisualsystemNeedstrainingNotautomatedIntrinsicbiasLimitofabout106or107observations(Wegman1995)Powerofintegrationwithanalyticalmethods6/25/20216DataMining:ConceptsandTechniquesScopeofVisualDataMiningVisual

6、ization:Useofcomputergraphicstocreatevisualimageswhichaidintheunderstandingofcomplex,oftenmassiverepresentationsofdataVisualDataMining:TheprocessofdiscoveringimplicitbutusefulknowledgefromlargedatasetsusingvisualizationtechniquesComputerGraphicsHighPerformanceCom

7、putingPatternRecognitionHumanComputerInterfacesMultimediaSystems6/25/20217DataMining:ConceptsandTechniquesPurposeofVisualizationGaininsightintoaninformationspacebymappingdataontographicalprimitivesProvidequalitativeoverviewoflargedatasetsSearchforpatterns,trends,

8、structure,irregularities,relationshipsamongdataHelpfindinterestingregionsandsuitableparametersforfurtherquantitativeanalysisProvideavisualproofofcomputerrepres