ChemometricsandIntelligentLaboratorySystems126(2013)11–20ContentslistsavailableatSciVerseScienceDirectChemometricsandIntelligentLaboratorySystemsjournalhomepage:www.elsevier.com/locate/chemolabModelNOxemissionsbyleastsquaressupportvectormachinewithtuningbasedonamelioratedteaching–learning-basedoptimizationGuoqiangLia,b,PeifengNiua,b,⁎,WeipingZhanga,b,YongchaoLiua,baKeyLabofIndustrialComputerControlEngineeringofHebeiProvince,YanshanUniversity,Qinhuangdao066004,ChinabNationalEngineeringResearchCenterforEquipmentandTechnologyofColdStripRolling,Qinhuangdao066004,ChinaarticleinfoabstractArticlehistory:Theteaching–learning-basedoptimization(TLBO)isanewefficientoptimizationalgorithm.ToimprovetheReceived14November2012solutionqualityandtoquickentheconvergencespeedandrunningtimeofTLBO,thispaperproposesanReceivedinrevisedform19April2013amelioratedTLBOcalledA-TLBOandtestitbyclassicalnumericalfunctionoptimizations.ComparedwithAccepted24April2013otherseveraloptimizationmethods,A-TLBOshowsbettersearchperformance.Inaddition,theA-TLBOisAvailableonline2May2013adoptedtoadjustthehyper-parametersofleastsquaressupportvectormachine(LS-SVM)inordertobuildNOxemissionsmodelofa330MWcoal-firedboilerandobtainawell-generalizedmodel.ExperimentalKeywords:Teaching–learning-basedoptimizationresultsshowthatthetunedLS-SVMmodelbyA-TLBOhaswellregressionprecisionandgeneralizationLeastsquaressupportvectormachineability.NOxemissions©2013ElsevierB.V.Allrightsreserved.Coal-firedboiler1.Introductiontoreplacetraditionalquadraticprogrammingmethod.ThesimplicityandinheritedadvantagesofSVMsuchasexcellentgeneralizationabili-Withtheincreaseofenergyconsumptionworldwideandim-tyandauniquesolutionpromotetheapplicationofLS-SVMinmanyprovedawarenessofenvironmentalprotection,boilercombustionpatternrecognitionandregressionproblems.optimizationproblemofpowerplantsattractstheattentionoftechni-TheregressionaccuracyandgeneralizationabilityofLS-SVMareex-calstaffsandmanagers.Theboilercombustionop