密度核估計中核函數的迭代算法及對最優(yōu)窗寬的研究.pdf

1、燕山大學碩士學位論文密度核估計中核函數的迭代算法及對最優(yōu)窗寬的研究姓名：王金然申請學位級別：碩士專業(yè)：運籌學與控制論指導教師：宋向東20051001燕山人學理學碩士學位論文Abstract’I’hedensitykernelestimation1Samel：hodofnon‘parameterstatisticsusedinthedistributionfunctioninsolvingstatisticalproblemswhichh

2、asimportantvalueofapplicationineconomicandfinancialfieldsAsfarasthiscontenttosaytheemphasisistostudyitsalgorithmofdensitykernelestimationSOthattheresultismoreaccurateThispaperwemaindiscussthefollowingthreecontentsFirst，w

3、ebeginwiththebasicknowledgeofdensitykernelestimationIncludingthebackgroundandthedefinitionofkernelestimationAccordingtotheintroductionofkernelfunctionandwindowwidth，GivingthegeneralsituationofkernelestimationAfterthat，we

4、studythebasicpropertieswithlargesamplesIntheseproperties，wehighlighttheestimationforthemodeofmultivariatedensityUsingthemethodofprobabilityandmeasurementprovethattheexistenceandtheexclusiveofthemodeShowingsometheoriesand

5、itsproofsSecond，itisabouttheproblemofhowtOchoosetheoptimalkernelfunction，InthesechapterBasedonthefourcommonkernelfunctions(parzen、cosinus、gaussandtrianglekernel)andfixedsampleswetrytosimulatethecomputationalprocesswhendi

6、fferentparameteranddifferentwindow—widthinusingMatlabthroughgraphicsandtables，AnalysisingtheinassdatagettheoptimalvalueofwindowwidthcorrespondingeverykernelfunctionAtlastanercomparingtheMSEandthegraphicsunderindividualwi

7、ndowwidthobtainingtheconclusionofoptimalkernelfunctionisthecosinuskernelThefollowingcontentsisthekeypointinallthepaperAtthebeginwebringaboutanewmethodoftheiterationalgorithm，Thischaptercoveringtheiterationalgorithmandsta

8、tisticalsimulateprocessofdensitykernelestimate，ComparingtheMSEstronglybeforeandafteriteration，thewholecomparingprocessanda11thedatashoWthattheideaofiterationisfeasible，Furthermore，Weobtainthetheorythattheiterationvalueis