Vol.3,No.4肖仕杰等:基于随机蛙跳和支持向量机的牛乳收购分级模型构建 85 [25]VapnikVN.Anoverviewofstatisticallearningtheo‐的柑橘黄龙病快速检测方法[J].智慧农业,2019,1 ry[J].IEEETransactionsonNeuralNetworks,1999,10(3):77-86. (10):988-999.DAIF,QIUZ,QIUQ,etal.Rapiddetectionofcitrus [26]BurgesCJC.ATutorialonsupportvectormachinesHuanglongbingusingRamanspectroscopyandauto- forpatternrecognition[J].DataMiningandKnowledgefluorescencespectroscopy[J].SmartAgriculture,2019, Discovery.1998,2(2):121-167.1(3):77-86. [27]BONFATTIV,MARTINOGD,CARNIERP.Effec‐[29]胡翼然,李杰庆,刘鸿高,等.基于支持向量机对云南 tivenessofmid-infraredspectroscopyforthepredic‐常见野生食用牛肝菌中红外光谱的种类鉴别[J].食 tionofdetailedproteincompositionandcontentsof品科学,2021,42(8):248-256. proteingeneticvariantsofindividualmilkofSimmen‐HUY,LIJ,LIUH,etal.Speciesidentificationofcom‐ talcows[J].JournalofDairyScience,2010,94(12):monwildedibleboleteinYunnanbyFouriertransform 5776-5785.mid-infraredspectroscopycoupledwithsupportvector [28]代芬,邱泽源,邱倩,等.基于拉曼光谱和自荧光光谱machine[J].FoodScience,2021,42(8):248-256. ConstructionofMilkPurchaseClassificationModelBasedon ShuffledFrogLeapingAlgorithmandSupportVectorMachine 11,23,444 XIAOShijie,WANGQiaohua,LIChunfang,ZHAOLimei,LIUXinya, 43 LUShiyu,ZHANGShujun (1.CollegeofEngineering,HuazhongAgriculturalUniversity,Wuhan430070,China;2.KeyLaboratoryofAgricul‐ turalEquipmentintheMid-LowerReachesoftheYangzeRiver,MinistryofAgricultureandRuralAffairs,Wuhan 430070,China;3.KeyLaboratoryofAnimalBreedingandReproductionofMinistryofEducation,HuazhongAgri‐ culturalUniversity,Wuhan430070,China;4.HebeiAnimalHusbandryAssociation,Shijiazhuang050031,China) Abstract:Protein,fatandsomaticcellsarethreeimportantreferenceindicatorsinmilkpurchase,whichdeterminethequality andpriceofmilk.Thetraditionalchemicalanalysismethodsoftheseindexesaretime-consumingandpollutetheenvironment, whilethemid-infraredspectrumhastheadvantagesoffast,non-destructiveandsimpleoperation.Inordertorealizetherapid classificationofmilkqualityandimprovetheproductionefficiencyofdairyenterprises,3216Holsteinmilksampleswerecho‐ senastheresearchobjectsandmid-infraredspectroscopytechnologywasappliedtorealizethedetectionandclassificationof4 differentqualitymilksduringthepurchaseprocess.Thespectrumwaspreprocessedbyusingthefirstderivativeandthefirstdif‐ ference,andcombinedwiththealgorithmcompetitiveadaptivereweightedsampling(CARS)andtheshuffledfrogleapingalgo‐ rithm(SFLA),theeffectivecharacteristicvariablesthatcouldrepresentdifferentmilkswereselected,andtheSVMmodelwas established.Amongthem,thepenaltyparametercandthekernelfunctionparametergwhichwerethekeyparametersofthe SVMmodelwereoptimizedbyusingthegridsearchmethod(GS),geneticalgorithm(GA)andparticleswarmalgorithm (PSO).ThetrainingtimeofGS,GAandPSOalgorithmswerecompared,theresultsshowedthatthetrainingtimeofGSwas muchlongerthanthatofGAandPSOalgorithms.TheSFLAalgorithmwasgenerallybetterthantheCARSalgorithm,andthe PSOoptimizedtheSVMmodelthebest.Afterthefirst-orderdifferencepreprocessing,thePSO-SVMestablishedbyusingthe SFLAalgorithmtofilterthecharacteristicvariables,theaccuracyofthetrainingset,theaccuracyofthetestsetandtheAUC were97.8%,95.6%and0.96489,respectively.Thismodelhasahighaccuracyrateandhaspracticalapplicationvalueinthe milkindustry. Keywords:mid-infraredspectrum;milk;purchaseclassification;shuffledfrogleapingalgorithm;supportvectormachine (登陆www.smartag.net.cn免费获取电子版全文) |
|