数据说明：[1,22]为自变量X，第23列为因变量Y
library(randomForest)library(ggplot2)library("caret")#当前工作路径getwd()#设置自己的工作路径，数据均可输出至该文件夹setwd('D:')data.all <- read.csv("test.csv")set.seed(9)folds <- createFolds(y=data.all$Y,k=10)max=0num=0for(i in 1:10){ fold_test <- data.all[folds[[i]],] #取folds[[i]]作为测试集 fold_train <- data.all[-folds[[i]],] # 剩下的数据作为训练集 print("***组号***") fold_pre <-randomForest(Y~.,data= fold_train,mtry=6,ntree=820, proximity=TRUE,importance=TRUE)fold_predict <- predict(fold_pre,type='response',newdata=fold_test) fold_predict =ifelse(fold_predict>0.5,1,0) fold_test$predict = fold_predict fold_error = fold_test[,24]-fold_test[,23] #Y（真实值）在23列，模型预测值在24列fold_accuracy = (nrow(fold_test)-sum(abs(fold_error)))/nrow(fold_test) print(i) print("***测试集精确度***") print(fold_accuracy) print("***训练集精确度***")fold_predict2 <- predict(fold_pre,type='response',newdata=fold_train) fold_predict2 =ifelse(fold_predict2>0.5,1,0) fold_train$predict = fold_predict2 fold_error2 = fold_train[,24]-fold_train[,23] fold_accuracy2 = (nrow(fold_train)-sum(abs(fold_error2)))/nrow(fold_train) print(fold_accuracy2) if(fold_accuracy>max) { max=fold_accuracy num=i } }print(max)print(num)#精度最高样本#测试testi <- data.all[folds[[num]],]#训练traini <- data.all[-folds[[num]],]#利用精度最高样本重新构建模型prei <- randomForest(Y~.,data= traini,ntree=820,mtry=6,proximity=TRUE,importance=TRUE)#看因子重要性importance<-importance(x= prei)importanceset.seed(100)varImpPlot(prei)#训练测试集predicti <- predict (prei,type='response',newdata=testi)testi$predict = predicti#训练训练集predicti2 <- predict (prei,type='response',newdata=traini)traini$predict = predicti2#训练全体predicti3 <- predict (prei,type='response',newdata=data.all)data.all$predict = predicti3#输出训练、测试、全体预测值（也可不用输出）write.csv(traini,"ausfold_train.csv")write.csv(testi,"ausfold_test.csv")write.csv(data.all,"ausfold_all.csv")#绘制ROC曲线library(ROCR)#训练： pred <- prediction(predicti2,traini$Y) #预测值(0.5二分类之前的预测值)和真实值 performance(pred,'auc')@y.values #AUC值perf <- performance(pred,'tpr','fpr')plot(perf,col="red")#测试：pred2 <- prediction(predicti,testi$Y) performance(pred2,'auc')@y.values perf2 <- performance(pred2,'tpr','fpr')plot(perf2,add=TRUE, col ="blue")#add即为加在上一个ROC图上面（训练）#全体：pred3 <- prediction(predicti3,data.all$Y) performance(pred3,'auc')@y.values perf3 <- performance(pred3,'tpr','fpr')plot(perf3,add=TRUE, col ="green")#混淆矩阵#训练：predict =ifelse(predicti2>0.5,1,0)traini$predict = predicttrue_value=traini[,23]predict_value=traini[,24]#计算模型精确度error = predict_value-true_valueaccuracy = (nrow(traini)-sum(abs(error)))/nrow(traini)precision=sum(true_value & predict_value)/sum(predict_value)recall=sum(predict_value & true_value)/sum(true_value)F_measure=2*precision*recall/(precision+recall) print(accuracy)print(precision)print(recall)print(F_measure)#混淆矩阵，显示结果依次为TP、FN、FP、TNtable(true_value,predict_value) #测试predict =ifelse(predicti>0.5,1,0)testi$predict = predicttrue_value=testi[,23]predict_value=testi[,24]#计算模型精确度error = predict_value-true_valueaccuracy = (nrow(testi)-sum(abs(error)))/nrow(testi)precision=sum(true_value & predict_value)/sum(predict_value)recall=sum(predict_value & true_value)/sum(true_value)F_measure=2*precision*recall/(precision+recall) print(accuracy)print(precision)print(recall)print(F_measure)#混淆矩阵，显示结果依次为TP、FN、FP、TNtable(true_value,predict_value) #全体：predict =ifelse(predicti3>0.5,1,0)data.all$predict = predicttrue_value=data.all[,23]predict_value=data.all[,24]#计算模型精确度error = predict_value-true_valueaccuracy = (nrow(data.all)-sum(abs(error)))/nrow(data.all)precision=sum(true_value & predict_value)/sum(predict_value)recall=sum(predict_value & true_value)/sum(true_value)F_measure=2*precision*recall/(precision+recall) print(accuracy)print(precision)print(recall)print(F_measure)#混淆矩阵，显示结果依次为TP、FN、FP、TNtable(true_value,predict_value)