Dual membership SVM method based on spectral clustering

在线阅读 下载PDF 导出详情
摘要 Anewfuzzysupportvectormachinealgorithmwithdualmembershipvaluesbasedonspectralclusteringmethodisproposedtoovercometheshortcomingofthenormalsupportvectormachinealgorithm,whichdividesthetrainingdatasetsintotwoabsolutelyexclusiveclassesinthebinaryclassification,ignoringthepossibilityof'overlapping'regionbetweenthetwotrainingclasses.Theproposedmethodhandlessample'overlap'efficientlywithspectralclustering,overcomingthedisadvantagesofover-fittingwell,andimprovingthedataminingefficiencygreatly.Simulationprovidesclearevidencestothenewmethod.
机构地区 不详
出版日期 2012年02月12日(中国期刊网平台首次上网日期,不代表论文的发表时间)
  • 相关文献