A novel local manifold-ranking based K-NN for modeling the regression一种新的基于局部流形排序的K-NN回归模型 生物活性与分子描述符之间的关系

《A novel local manifold-ranking based K-NN for modeling the regression一种新的基于局部流形排序的K-NN回归模型 生物活性与分子描述符之间的关系》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、ChemometricsandIntelligentLaboratorySystems151(2016)71–77ContentslistsavailableatScienceDirectChemometricsandIntelligentLaboratorySystemsjournalhomepage:www.elsevier.com/locate/chemolabAnovellocalmanifold-rankingbasedK-NNformodelingtheregressionbetweenbioactivityandmolec

2、ulardescriptorsa,1b,1c,⁎,XinHuangdcdLiangShen,DongshengCao,QingsongXu,NanXiao,YizengLiangaSchoolofScience,QingdaoUniversityofTechnology,Qingdao266520,PRChinabSchoolofMedicine,CentralSouthUniversity,Changsha410083,PRChinacSchoolofMathematicalandStatistics,CentralSouthUniv

3、ersity,Changsha410083,PRChinadResearchCenterofModernizationofTraditionalChineseMedicines,CentralSouthUniversity,Changsha410083,PRChinaarticleinfoabstractArticlehistory:Inthepresentstudy,weproposeanovellocalregressionalgorithmbasedonmanifold-rankingandk-nearestReceived10F

4、ebruary2015neighbors(MRKNNforshort).Undertheframeworkofkernelmethods,thegrouprelationshipsharedamongReceivedinrevisedform9December2015multiplemoleculesisfirstlycapturedbythegraphwherenodesrepresentmoleculesandedgesrepresentAccepted10December2015pairwiserelations.Then,mani

5、foldrankingalgorithmisdevelopedforquery-orientedextractivesummarization,Availableonline21December2015wheretheinfluenceofqueryispropagatedtoothermoleculesthroughthestructureoftheconstructedgraph.WhenevaluatedonfourSARdatasets,MRKNNalgorithmcanprovideafeasiblewaytoexploitth

6、eintrinsicstruc-Keywords:Chemicalsimilaritytureofsimilarityrelationships.Resultshavevalidatedtheefficacyoftheproposedalgorithm.Manifold-ranking©2015ElsevierB.V.Allrightsreserved.k-nearestneighbors(K-NN)Quantitativestructure–activityrelationship(QSAR)1.Introductionsubsets.

7、Actually,oneofthemostcommonlyusedlocalmethodsinQSARstudyingisk-nearestneighbors(K-NN)[18].ItisoneoftheQuantitativestructure–activityrelationship(QSAR)technologymostfundamentalandsimpleclassificationorregressionmethodsforiscapableofmodelingandpredictionoftherelationshipbet

8、weenaQSARstudywhenthereislittleornopriorknowledgeaboutthedistri-response-variableandmolecularpredictors