高级分类器：支持向量机( SVM).rar_R svm_SVM_SVM.R

#加载包 library(e1071) help(package="e1071") #加载数据集 setwd("D:\\R_edu\\data") orgData<-read.csv("date_data2.csv") #查看是否存在缺失值 summary(orgData) #提取如下字段进行建模 orgData1=orgData[,c("income","attractive","assets","edueduclass","Dated")] orgData1$edueduclass=as.factor(orgData1$edueduclass) orgData1$Dated=as.factor(orgData1$Dated) #构建训练集和测试集 set.seed(110) select<-sample(1:nrow(orgData1),length(orgData1$Dated)*0.6) train=orgData1[select,] test=orgData1[-select,] train.y=orgData1[select,5] test.y=orgData1[-select,5] #使用svm函数，建立支持向量机模型 svm.mod<-svm(Dated~income+attractive+assets+edueduclass,kernel="radial", data=train,probability=TRUE,cost=0.5,gamma=0.4,cross=10) summary(svm.mod) #画出在svm模型中，两个变量的关系图 svm.mod<-svm(Dated~attractive+assets,kernel="radial", data=train,probability=TRUE,cost=0.5,gamma=0.4,cross=10) plot(x=svm.mod,data=train,formula=assets~attractive,svSymbol = "x", dataSymbol = "o") plot(x=svm.mod,data=train,formula=assets~income,svSymbol = "x", dataSymbol = "o") plot(x=svm.mod,data=train,formula=income~attractive,svSymbol = "x", dataSymbol = "o") ?plot.svm #在新数据做预测 y_hat=predict(svm.mod,orgData1,probability=F) #召回率和精确度 require(gmodels) t<-CrossTable(x =orgData1$Dated, y = y_hat,prop.chisq=FALSE) t$prop.row[2,2]#召回率 t$prop.col[2,2]#精确度 ######################################################################################## #选择最优模型参数 require(gmodels) ROC<-data.frame() for (i in seq(from =0.1,to =10,by =0.5)){ svm.mod<-svm(Dated~income+attractive+assets+edueduclass,kernel="polynomial", data=train,probability=TRUE,cost=i,gamma=0.4,cross=10) y_hat=predict(svm.mod,test,probability=F) t<-CrossTable(x =test.y, y = y_hat,prop.chisq=FALSE) accuracy<-sum(y_hat==test.y)/length(test.y)#准确率 out<-data.frame(i,accuracy,t$prop.row[2,2],t$prop.col[2,2]) ROC<-rbind(ROC,out) } names(ROC)<-c("n","accuracy","Recall","Precision")