向数据挖掘的遗传算法的研究与应用

2007262007262007228-1-HBIOLAPHBIOLAP-2-AbstractInBusinessIntelligenceSystemthetraditionalclusteranalysishasbeenwidelyused,buttherearestillsomeproblems.Whendealingwithlargedataconvergence,forexample,itwillsometimeslowerandmorepronetolocalminimumproblems.Inviewoftheseproblems,thepaperpresentsaclusteranalysismethodbasedontheimprovedgeneticalgorithms,anapplication-orientedsmallandmedium-sizedenterprisesandbusinessintelligencesystemforHBI.Thedesignandimplementationofonlineanalyticalprocessingsubsystems,andexperimentsshowthattheclusteringalgorithmisbetterqualityandperformance.ThemajorworkinthispaperlistsasfollowsFirst,thepapersdetail-orientedbusinessintelligencesystemsandrelatedtechnologyClusteranalysisoftheexistingtechnologies.Second,inresponsetotheslowconvergence,andclusteranalysisofexistingearly,aclusteranalysisbasedontheimprovedgeneticalgorithms,includingchromosomecodingmeetthetermscrossoverandmutation.Meanwhiletheperformanceofthisalgorithmarecomparedwithotherclusteranalysisalgorithm,theexperimentalresultsshowthattheclusteringalgorithmisbetterqualityandperformance.Finally,describedthebusiness-orientedsmallandmedium-sizedenterprisesintheonlineanalyticalprocessingsubsystemHBIIntelligent-3-SystemDesignandImplementation,Thesystemincludeadatawarehousebottom-lineanalysisandclusteranalysisplatformmodule.Supermarketsalessystemandthespecificsceneofsupermarketsalesdatamining,theresultsshowthatthegeneticalgorithmbasedonimprovedsalesoftheclusteranalysisinthesupermarketsystemhavemadethepracticalapplicationofagoodclustereffect.Keywords:Datamining;Clustering;OLAP;Geneticalgorithms;ImprovedGeneticalgorithms;BusinessIntelligence-6-1.1(BI)1.2BIBIBusinessObjects,IBM,Informix,Microsoft,OracleCA,SAS,Congo’s,SybaseNCRCA-7-1-1Fig1-1ThebasicarchitecturediagramofBusinessIntelligenceSystemCACAODBC(Oracle,Sybase,SQLServer,InformixIBMDB2)Erwin,DecisionBase,InfoPump,InfoBeacon,ForestTrees,Repository,ProVisionaCognosBusinessObjectsOLAPWindowsLinux,UnixCognosImpromptuScenario(OLAP)PowerPlayerBusinessObjectsBusinessObjectsWebWebintelligenceBusinessminer,BusinessminerIBM,Oracle,Informix,Sybase,NCR,SAS,Microsoft.OracleOLAPOracleSQLOracleOracle81OracleWarehouseBuilder,(1)Olap(2)(3)(4)(5)-8-OracleDeveloperServerClient/ServerWebOracleDiscovererWEBOracleDarwinIBM(B)SQLIBMBusinessObjectsBusinessObjectsSASSASIBMVisualWarehouse,Essbase/DB2OLAPServer,QUESTSybaseBitjisc100SybaseIndustryWarehouseStudioWarehouseStudioWarehouseControlCenterIInformixFastStartROLAPInformixIDSIDS/ADInformixMicrosoftIISNetscapeEnterprise/FastTrackwebInformix(BriSAS)InformixInformixInformixNCR(ScalableDataWarehouse,SDW)oNCRNCRNCRTeradataNCRWorldMarkSMPUnix(DataMart)1998WindowsNTTeradataoSASSASSASSASLicenseSAS/WA(WarehouseAdministrator)SAS/MDDB,SAS/AF,SAS/ITSV(ITServiceVision)Microsoft-9-Microsoft1.31);;2)3)GAGAGAGAGAGA4)1.420(cluster),-10-WEB(CRM);;;;1.51);;2)1.6(CustomerRelationmanagementCRM)-11--12-2.12.1.1[1](DataMining-DM)1995(1)(2)(3)(4)()-13-.OLAP?;/OLAP?OLAP?OLAPOLAP/;OLAP;OLAPOLAP()OLAPOLAPitemOLAPOLAMOLAPOLAP2.1.2KDDKDDKDD-14-()2.1.3[33]1.2.3.4.5.6.7.8.9.2-5Fig2-5Dataminingprocess-15-2.1.412[1112]90%(++)3[1314](classification)()()[15]Bayes[1617]k-[1][18][19](prediction)4(partitioningmethod)(hierarchicalmethod)(density-basedmethod)(grid-basedmethod)(model-basedmethod)5[2021](summarized)(concise)(precise)(conceptdescription)6[2223]-16-2.1.51.562.70DBMSSQL(I);(2);(3);(4)Web;(S);(6);(7);(8)WebDMKD-17-DMKD2.2(Clustering)(cluster),WEB(clustering)(cluster)2.2.13R-18-(numerical)(binary),(categoricalnominal),(ordinal)(Euclideandistance)(Manhattandistance)()2.2.2(partitioningmethod)(partitioningmethod)MKK=M.KK-19-1)k-2)k-(hierarchicalmethod)(hierarchicalmethod)()();1)CUERChameleon2)BIRTH(density-basedmethod)(density-basedmethod)()DBSCANOPTICS(grid-basedmethod)(grid-basedmethod)STINGCLIQUEWaveCluster(model-basedmethod)(model-basedmethod)2.3-20-1)2)nO(n)3)4)5)2.3.11)fF-q.;0,I)Gray2)3)-21-(selection)(crossover)(mutation)NN1);(0,1)Gray2.3.2;4-1(SGA)SteplnPCFm,SC;Step2m;Step3t=0;Step4;Steps;Step6;Step7Set=t+l,Step4;SeSc-22-4-1Fig4-1thebasicprocessofgeneticalgorithms2.42.4.1(BusinessIntelligence)-23-MIS(ManagementInformationSystem)MISMISMISMISMISMISMISMIS(1);(2);(3)OLAPOLAP2.4.2(DataWarehouse),(OLAP-OnlineAnalysisandProcess)(DataMining).(DW)W.H.Inmon(OLAP)(DM)AI-24-IGA(1)(2)(3);(4)(5)(6)(7);.(8)(9)-25-3.13.2GA3.3IGA;;BeginInitializeX,cNPrPmmaxgen;;P(0);ForI=1tomaxgendo(I-1);P(i-1)P(I);Endfor-26-3.3.1;Cvi(i=1,2,...,c)chr=v1v2v3…C*X(l}U(2)(1-2)V(3)NN3.3.2;(FitnessFunction)-27-JJi;3.3.3;()()(1)(2)-28-1)2)3)(01)(3)3.3.4N}lVl2}(W/2}Nl2)()chrl,chr2α;-29-A,BA',B'3-13-1Fig3-1informationoverlappingandlossdirectlycausedbytheSingle-pointcross3.3.5{0,1}0110[]minmax,uu;01;1O-30-3.4GAGAk-means,GA-means,IGA;Iris()IrisIris4350150n=100pc=0.60pm=0.01,T=100.Wine()178133n=200pc=0.60pm=0.01T=1000.Automotives()61211n=200pc=0.60pm=0.02T=1000k-means,GA-means,IGA3-2203-2Irisk-meansGA-meansIrisWinek-meansWinek-meansGA-means3-2Fig3-2ComparisonofThreeclusteringalgorithm-31-4-3Fig4-3Effectofthethreealgorithmto