第41卷第3期陕西科技大学学报Vol.41No.32023年6月JournalofShaanxiUniversityofScience&TechnologyJun.2023*文章编号:2096-398X(2023)03-0208-07基于核KMeans和SOM神经网络算法的海况聚类分析陈晓曼,苏欢*(哈尔滨工业大学(威海)理学院,山东威海264209)摘要:为了更加高质量地利用海况数据,避免由复杂因素导致的对海况误判问题,基于KMeans、核技巧、自组织映射(Self-organizingMapping,SOM)神经网络构建了自组织映射混合核KMeans(SOM-GaussianandPolynomialKernel-KMeans,SGPK-KMeans)算法.克服了KMeans对复杂数据聚类效果不佳、核KMeans需要指定聚类数目和对初始聚类中心敏感的问题.通过海况数据聚类实验,将SGPK-KMeans算法的聚类效果与经典KMeans、单核KMeans和SOM神经网络算法进行对比分析.结果表明SGPK-KMeans对于海况数据聚类具有更加稳定的效果且能更加准确的识别出数据中的异常值.关键词:聚类;海况;核KMeans;SOM神经网络中图分类号:TP391文献标志码:ASeastateclusteringanalysisbasedonkernelKMeansandSOMneuralnetworkalgorithmCHENXiao-man,SUHuan*(SchoolofScience,HarbinInstituteofTechnology(Weihai),Weihai264209,China)Abstract:Inordertomakebetteruseofseastatedataandavoidmisjudgmentofseastatecausedbycomplexfactors,Self-OrganizingMappingmixedkernelKMeans(SOM-GaussianandPolynomialKernel-KMeans,SGPK-KMeans)algorithmhasbeenconstrustedontheba-sisofKMeans,kernelskillsandSelf-organizingMapping(SOM)neuralnetworkforcomplexseastatedata.Itovercomesthefollowingproblems,forexample,KMeanshaspoorclusteringeffectoncomplexdata,kernelKMeansneedstospecifythenumberofclustersandissensi-tivetotheinitialclusteringcenter.Withtheseastatedataclusteringexperiment,thecluste-ringeffectofSGPK-KMeansalgorithmiscomparedwiththatofclassicalKMeans,single-coreKMeansandSOMneuralnetworkalgorithms.ThefindingsshowthatSGPK-KMeanshasamorestableeffectonseastatedataclusteringandcanidentifyoutliersinthedatamoreaccu-rately.Keywords:clustering;seastate;kernelKMeans;SOMneuralnetwork*收稿日期:2023-01-22基金项目:山东省自然科学基金面上项目(ZR202102220411)作者简介:陈晓曼(1998—),女,内蒙古呼伦贝尔人,在读硕士研究生,研究方向:数据分析通讯作者:苏欢(1981—),女,黑龙江佳木斯人,副教授,博士生导师,...
