人工智能方法在色谱保留值中的应用,综述


AnalyticaChimicaActa537(2005)

101–110
Predictionofretentiontimeofavarietyofvolatileorganiccompounds
basedontheheuristicmethodandsupportvectormachine
FengLuana,ChunxiaXuea,RuishengZhanga,b, ,ChunyanZhaoa,
MancangLiua,ZhideHua,BotaoFanc
c
DepartmentofChemistry,LanzhouUniversity,Lanzhou730000,ChinaDepartmentofComputerScience,LanzhouUniversity,Lanzhou730000,ChinaUniversit´eParis7-DenisDiderot,ITODYS1,RueGuydelaBrosse,75005Paris,France
b
a
Received14November2004;receivedinrevisedform23December2004;accepted23December2004
Availableonline29January2005
Abstract
Thesupportvectormachine(SVM)andtheheuristicmethod(HM)wereusedtodevelopthenon-linearandlinearmodelsbetweentheretentiontime(RT)and vemoleculardescriptorsof149volatileorganiccompounds(VOCs).Descriptorscalculatedfromthemolecularstructuresalonewereusedtorepresentthecharacteristicsofcompounds.The vemoleculardescriptorsselectedbytheheuristicmethodinCODESSAwereusedasinputsforSVM.TheresultsobtainedbySVMwerecomparedwiththoseobtainedbytheheuristicmethod.Themeansquarederrors(MSE)inRTpredictionsforthewholedatasetgivenbyHMandSVMwere1.270and0.844,respectively,whichshowstheperformanceofSVMmodelisbetterthanthatoftheHMmodel.Thepredictionresultsareinverygoodagreementwiththeexperimentalvalues.Thispaperprovidedanewandeffectivemethodforpredictingthechromatographyretentionindex.©2005ElsevierB.V.Allrightsreserved.
Keywords:QSRR;SVM;VOCs;Retentiontime
1.Introduction
Volatileorganiccompounds(VOCs)arecommoncon-stituentsofurbanair[1–4],andmany,suchasbenzene,styrene,andtoluene,areknownorsuspectedtocausechronicadversehealtheffectsinexposedpopulations[5].VOCsalsoplayamajorroleintheformationofvarioussecondarypollutantsthroughphotochemicalreactionsinthepresenceofsunlightandnitrogenoxides[6,7].Furthermore,someVOCscouldcontributetotheatmosphericozonedepletionandthebuild-uppersistentpollutionsinremoteareas[8,9].Therefore,thesecompoundshavebeenanimportanten-vironmentalissueoverthelasttwodecadesandhaveat-tractedsigni cantattentionfromdifferentresearchgroups.Therearemanyexperimentaltechniquedevelopedtoiden-
Correspondingauthor.Tel.:+869318912578;fax:+869318912582.E-mailaddress:ruison@public.lz.gs.cn(R.Zhang).
tifyandquantifythevolatileorganiccompounds[10–13].Amongthenumerousseparationtechniques,gaschromatog-raphyplaysamajorrole.However,therearesomelimi-tationsforthesemethods:thelimitationofsorbent-basedmethodaresorbents,whicharenormallyselectivein,ifnotlimitedto,adsorbing/absorbingcertainclassesofVOCs[14];thecanister-basedtechniqueisthoughttobeunsuit-ableforcollectingpolarcompounds;highsamplevolumesofhumidaircausedshiftsinthechromatographicelu-tiontimesthatcandegradeidenti cationandquanti ca-tionofnon-speci cdetectors[15].Besidestheabovemen-tioned,theexperimentaldeterminationofretentiontimeistime-consumingandexpensive.Alternatively,quantitativestructure–retentionrelationship(QSRR)providesapromis-ingmethodfortheestimationofretentiontimebasedondescriptorsderivedsolelyfromthemolecularstructureto texperimentaldata.Theadvantagesofthisapproachlieinthefactthatitrequiresonlytheknowledgeofchemical
0003-2670/$–seefrontmatter©2005ElsevierB.V.Allrightsreserved.doi:10.1016/j.aca.2004.12.085
基于启发式算法和支持向量机算法预测醛类化合物急性毒性_专业资料。应用启发式算法(HM)和支持向量机算法(SVM)建立了88种醛类化合物对大鼠急性毒性的定量构效关系...
基于启发式算法的混沌支持向量机流量预测_专业资料。针对现有混沌支持向量机回归模型存在流量预测效率低下的问题,利用差分进化(DE)算法、遗传算法和粒子群优化算法...
实验2分类预测模型_支持向量机_计算机软件及应用_IT/计算机_专业资料。简绍...虽然有些启发式的算法 可以获取这些参数值,但都不是最理想的。其中,正则化参数...
基于支持向量机的预测控制算法_专业资料。采用基于支持向量机的模型预测控制法来...其具体实现通过计算种群适应值函数、检查每个解是否满足约束条件、选取一定数量产生...
基于在线支持向量机和遗传算法的预测控制_机械/仪表_工程科技_专业资料 暂无评价
支持向量机算法用于烷基苯的热容和标准焓值的预测_专业资料。建立了有关烷基苯结构和它们的热容、标准焓值之间的定量构效关系模型。根据烷基苯的分子结构式,其每个...
一类基于支持向量机的软件故障预测方法_自然科学_...核函数取较大的值, 其性 能类 似多项 式核 函数...2. 3 基于 SVM 的软件故障预测原理及算法 软件...
启发式搜索算法首先随机选择一个起始点, 然 后通过...地一部分形成新的 个体, 其目的是保存高适应度的...支持向量机对金融数 据进行预测, 在训练支持向量机...
(x, w)对y进 行预测时造成的损失,称为损失函数...在VC维和学习一致理论基础上,Vapnik等人证明了对二值...? ? SMO算法对线性支持向量机最为有效,对非线性则...
18 No. 7 July, 2006 支持向量机的时间序列回归与预测董 辉,傅鹤林,冷伍明...支持向量机回归算法:给定 k 个样本数据,其值表示 为:{x k , y k } ,式...
热门文档