資源描述：
《knowledge management》由會員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、KnowledgeManagement----文化教育論文-->IntroductionPrincipalponentanalysisisakindofdiversestatisticsanalysismethod.Principalponentanalysisisalsoatechniqueusedtoreducemultidimensionaldatasetstoloensionsforanalysis.PCAostlyusedasatoolinexploratorydataanalysisandformakingpredictivemo
2、dels.PCAinvolvesthecalculationoftheeigenvaluedepositionofadatacovariancematrixorsingularvaluedepositionofadatamatrix,usuallyaftermeancenteringthedataforeachattribute.TheresultsofaPCAareusuallydiscussedintermsofponentscoresandloadings.Foroverallanalyzeprobleminactualtopic,pe
3、oplesusuallyputforeinformationatdissimilaritydegree.Itselectsimportantvariablethroughlinetransformationofmanyvariables.Butethodresearchthemultivariatestatisticalanalysistopic,toomanyvariablesoreinformationthroughfeationhascertaindegreeduplicationinthistopic.Principleponenta
4、nalysisestablishespossiblyfeakesteantimetheseneationintheaspectoftopicinformationreflection.Principleponentanalysisgetsaneetimeaccordingtoeffectivedemand,itcantakeoutseveraltotalvariables.Thesetotalvariablescanreflectoriginalvariables’informationpossibly.Principleponentanal
5、ysisisalsoamathematicmethodfordealathematicallydefinedasanorthogonallineartransformationthattransformsthedatatoanesuchthatthegreatestvariancebyanyprojectionofthedataestolieonthefirstcoordinate(calledthefirstprincipalponent),thesecondgreatestvarianceonthesecondcoordinate,and
6、soon.PCAistheoreticallytheoptimumtransformforagivendatainleastsquareterms.PCAcanbeusedfordimensionalityreductioninadatasetbyretainingthosecharacteristicsofthedatasetthatcontributemosttoitsvariance,bykeepingloostimportant"aspectsofthedata.HoaynotalathematicalprocessingofPCAm
7、eansusealinearitybinationoforiginalPpiecesindextodoan-->eethodisusevarianceofF1(thefirstlinearbination)todoaexpression.ItmeansbiggervariancerepresentsF1containsmoreinformation.ThereforeF1varianceselectedinalloflinearbinationsshouldbebiggest.SocalledF1isfirstprincipleponent.
8、IfthefirstprincipleponentisnotenoughforrepresentationofPpiecesofindexinformation,t
