R语言进行集成学习算法：随机森林

卡卡_R-Python2024-07-16 8:45
R 复制代码
# 10.4 集成学习及随机森林
# 导入car数据集
car <- read.table("data/car.data",sep = ",")
# 对变量重命名
colnames(car) <- c("buy","main","doors","capacity",
                   "lug_boot","safety","accept")
# 随机选取75%的数据作为训练集建立模型，25%的数据作为测试集用来验证模型
library(caret)
library(ggplot2)
library(lattice)
# 构建训练集的下标集
ind <- createDataPartition(car$accept,times=1,p=0.75,list=FALSE) 
# 构建测试集数据好训练集数据
carTR <- car[ind,]
carTE <- car[-ind,]
carTR<- within(carTR,accept <- factor(accept,levels=c("unacc","acc","good","vgood")))
carTE<- within(carTE,accept <- factor(accept,levels=c("unacc","acc","good","vgood")))


# 使用adabag包中的bagging函数实现bagging算法
#install.packages("adabag")
library(adabag)
bagging.model <- bagging(accept~.,data=carTR)

# 使用adabag包中的boosting函数实现boosting算法
boosting.model <- boosting(accept~.,data=carTR)

# 使用randomForest包中的randomForest函数实现随机森林算法
#install.packages("randomForest")
library(randomForest)
randomForest.model <- randomForest(accept~.,data=carTR,ntree=500,mtry=3)

# 预测结果，并构建混淆矩阵，查看准确率
# 构建result，存放预测结果
result <- data.frame(arithmetic=c("bagging","boosting","随机森林"),
                     errTR=rep(0,3),errTE=rep(0,3))
for(i in 1:3){
  # 预测结果
  carTR_predict <- predict(switch(i,bagging.model,boosting.model,randomForest.model),
                           newdata=carTR) # 训练集数据
  carTE_predict <- predict(switch(i,bagging.model,boosting.model,randomForest.model),
                           newdata=carTE) # 测试集数据
  # 构建混淆矩阵
  tableTR <- table(actual=carTR$accept,
                   predict=switch(i,carTR_predict$class,carTR_predict$class,carTR_predict))
  tableTE <- table(actual=carTE$accept,
                   predict=switch(i,carTE_predict$class,carTE_predict$class,carTE_predict))
  # 计算误差率
  result[i,2] <- paste0(round((sum(tableTR)-sum(diag(tableTR)))*100/sum(tableTR),
                              2),"%")
  result[i,3] <- paste0(round((sum(tableTE)-sum(diag(tableTE)))*100/sum(tableTE),
                              2),"%")
}
# 查看结果
result