Win.s with a clear geographical origin.are an.issue of in.erest for con.umers an. food *** paper presen.s a data min.n. study of Merlot win.s from South America to iden.ify the fin.erprin. of their geographical origin...
Win.s with a clear geographical origin.are an.issue of in.erest for con.umers an. food *** paper presen.s a data min.n. study of Merlot win.s from South America to iden.ify the fin.erprin. of their geographical origin.A group of samples from Argen.in.(n.17),Brazil(n.12),Chile(n.48),an. Uruguay(n.6)was *** chemical compoun.s were determin.d by high-performan.e liquid chromatography(HPLC).These compoun.s in.lude an.ioxidan. activity,total polyphen.ls,total an.hocyan.n.,in.ividual an.hocyan.n. an. *** binary classification.problems were performed(Brazil versus n.n.Brazil,Argen.in. versus n.n.Argen.in.,Chile versus n.n.Chile,an. Uruguay versus n.n.Uruguay)to in.estigate the geographic characteristics of each *** the evaluation.of binary classification. in.our dataset it was possible to iden.ify the main.variables(chemical compoun.s)that discrimin.te between.the *** used the followin. algorithms:Syn.hetic Min.rity over-sample Techn.que an. un.er-samplin. to balan.e the dataset of each classification.approach,the Relief algorithm to obtain.a variable importan.e ran.in. an. the classifiers Support Vector Machin.s,Multilayer Perceptron.an. Radial Basis Fun.tion.n.twork with dyn.mic decay *** model obtain.d the highest performan.e measures amon. the classifiers for each dataset(93.73%of accuracy for the Brazil versus n.n.Brazil,91.18%for the Argen.in. versus n.n.Argen.in.,79.16%for the Chile versus n.n.Chile,an. 91.67%for the Uruguay versus n.n.Uruguay classification..These accuracies were achieved by the search of the possible variable subsets accordin. to Relief for each classification.*** foun. that some variables,such as DPPH,win. color an. in.ividual an.hocyan.n.,are amon. the most importan. variables in.the characterization.of Merlot win.s.
The purpose of this paper is to classify win.s from 4 differen. coun.ries in.South *** class of win.s is formed by samples con.idered by experts as represen.atives of the followin. commercial categories:“Argen.in.an....
详细信息
The purpose of this paper is to classify win.s from 4 differen. coun.ries in.South *** class of win.s is formed by samples con.idered by experts as represen.atives of the followin. commercial categories:“Argen.in.an.Malbec(AM)”,“Brazilian.Merlot(BM)”,“Uruguayan.Tan.at(UT)”an.“Chilean.Carmén.re(CC)”.The 83 samples collected were an.lyzed accordin. to their composition.of volatiles,semi-volatiles an. phen.lic *** built a decision.system for classification.based on.support vector machin.s(SVM),alon. with Correlation.based Feature selection.CFS),an. Ran.omForest Importan.e(RFI),whichmeasures the relative importan.e of the in.ut ***,we use CFS to select a subset of variables amon. 190 chemical *** chemicals were selected as correlated to the category an. un.orrelated with each ***,these chemical compoun.s were organ.zed accordin. to the importan.e ran.in. given.by the RFI an. classified with *** study clearly in.icated that SVMin.combin.tion.with feature selection.methodswas able to iden.ify the most importan. chemicals to classify the win. *** the compoun.s iden.ified in.the win. samples,the variable subset defin.d by the feature selection.methods,which were catechin.gallic,octan.ic acid,myricetin.caffeic,isobutan.l,resveratrol,kaempferol,an. ORAC,were able to achieve an.accuracy of 93.97%in.classifyin. the commercial categories.
暂无评论