%0 Journal Article
%T 自组织映射算法与基于专家系统的支持向量回归的结合
%A 王玲
%A 穆志纯
%A 郭辉
%J 自动化学报
%P 612-619
%D 2005
%X ？AnewapproachisproposedtomodelnonlineardynamicsystemsbycombiningSOM(self-organizingfeaturemap)withsupportvectorregression(SVR)basedonexpertsystem.Thewholesystemhasatwo-stageneuralnetworkarchitecture.InthefirststageSOMisusedasaclusteringalgorithmtopartitionthewholeinputspaceintoseveraldisjointedregions.Ahierarchicalarchitectureisadoptedinthepartitiontoavoidtheproblemofpredeterminingthenumberofpartitionedregions.Then,inthesecondstage,multipleSVR,alsocalledSVRexperts,thatbestfiteachpartitionedregionbythecombinationofdifferentkernelfunctionofSVRandpromotetheconfigurationandtuningofSVR.Finally,toapplythisnewapproachtotime-seriespredictionproblemsbasedontheMackey-GlassdifferentialequationandSantaFedata,theresultsshowthatSVRexpertshaseffectiveimprovementinthegeneralizationperformanceincomparisonwiththesingleSVRmodel.
%K SOMclustering
%K SVRexperts
%K singleSVR
%K Mackey-Glassdifferentialequation
%K SantaFedata
%U http://www.aas.net.cn/CN/abstract/abstract16028.shtml