基于数据挖掘技术的决策支持系统及其关键技术研究

《基于数据挖掘技术的决策支持系统及其关键技术研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、南京理工大学博士学位论文基于数据挖掘技术的决策支持系统及其关键技术研究姓名：汤九斌申请学位级别：博士专业：模式识别与智能系统指导教师：杨静宇20090101摘要博：十：论文型库管理更加方便。关键词：决策支持系统，电信，数据仓库，数据挖掘，聚类，EPNET预测，层次分析法，模型库，模型管理系统U博：l：论文基于数据挖掘技术的决策支持系统及其关键技术研究AbstractWithChina’SsuccessfulentryintoWTO，thegiobalizationofworldeconomyhascausedTelecomCorporationtofaceunprece

2、dentedchallenge．RapidandscientificdecisionconcernsthefateofTelecomCorporationtoagreatextent．Atpresent，Telecomoperatorsgenerallyadoptthetraditionaldecision·makingway,whichisanalyzingdatafrommanualreports．Alotofproblemsresultfromthisway,suchashugeworkload,slownessandbeingeasytolosethedatae

3、re．Therefore，ithasbecomethecommonunderstandingtostudyanddeveloptheDecisionSupportSystem(DSS)adaptedtoChineseTelecomCorporationamongalltheTelecomoperators．Thispaperfocusesonthestudyofthedata-mining-baseddecisionsupportsystemanditskeytechniques，whichisthecoreoftheprojectof‘'TelecomIPDecisi

4、onSupportSystem(IPDSS)”．Atelecomdecisionsupportsystemsolutionintegratingdatawarehouseanddataminingtechniquesisdesigned，andthekeytechniquesusedinthe“IPDSS”，suchasclustering,predictionandthemodelingtechnology,areanalyzed．AimingatthefeatureandrequirementofIPservicesoftelecommunicationoperat

5、ors，adecision—makingsupportsystem(DSS)basedonthe．Netframeworkisdesigned．ThissystemadoptsASENETasthemodeloftheWebapplication，ADO．NETisusedasthefoundationofdataaccess，andSQLserver2000isasasolutionofdatabaseanddatawarehouse．Thetechnologyofdataminingandartificialintelligenceisembeddedeffecti

6、vely．Itcanprovidequick，scientificreferenceandfoundationfordecision-makingoftelecomIPoperation．K-Meansalgorithmisoneofcommonclusteringalgorithms，buttheclustercenterinitializationisahardproblem．Inthispaper,ahierarchical·basedinitializationapproachisproposedforK-Meansalgorithm．Thegeneralclu

7、steringproblemistreatedasweightedclusteringproblem，andtheoriginaldataissampledlevelbyleveltoreducethedataamount．Thenclusteringiscarriedoutateachlevelbytop-down．TheinitialCenterofeachlevelismappedfromtheclusteringofupperlevelandrepeatthisprocedureuntiltheoriginaldataleveli