使用 rocr 包的决策树的 ROC 曲线

Question

我使用 rpart 包来开发我的树并预测模型。最后为了绘制 ROC 曲线，我尝试使用 rocr 包。抱歉不能用内置数据集复制它。请找到我使用的 csv 的 link：

现在请看我的代码：

#setting up data
data<- read.csv(file.choose())
quality_binary <- ifelse(wine_quality >5,"high","low")
data <- data.frame(data,quality_binary)

#re shuffling the data
set.seed(9850)
g <- runif(nrow(data))
datar<- data[order(g),]
#removing the wine quality column since it has to be predicted
datar <- datar[-12]

library(rpart)  
library(rpart.plot)
library(cvTools) 
library(caret)
library(tree)

k <- 10 # setting the value for 10 fold validation 

folds <- cvFolds(NROW(datar), K=k)
datar$holdoutpred <- rep(0,nrow(datar))

for(i in 1:k){

train <- datar[folds$subsets[folds$which != i], ] #training set
validation <- datar[folds$subsets[folds$which == i], ] #validation set

#tree model
tree_model_rpart_gini = rpart(quality_binary~.,data = train,
        parms = list(split = "information"), method  = "class")
rpart.plot(tree_model_rpart_gini,type = 3,extra = 101)

#prediction
pred_model_rpart_gini <- predict(tree_model_rpart_gini,   
newdata=validation, type="class")

datar[folds$subsets[folds$which == i], ]$holdoutpred <-   
pred_model_rpart_gini

}

#plotting ROC curve

library(ROCR)
 pred1 <- prediction(predict(datar$pred_model_rpart_gini),   
 datar$quality_binary)
 perf1 <- performance(pred1,"tpr","fpr")
 plot(perf1)

而我的错误是：

pred1 <- prediction(predict(datar$pred_model_rpart_gini),   
datar$quality_binary)
Error in UseMethod("predict") : 
no applicable method for 'predict' applied to an object of class "NULL"

Answer 1

datar$pred_model_rpart_gini 是 NULL 即未定义。

你可能打算用 pred_model_rpart_gini（不是 datar$）代替？

使用 rocr 包的决策树的 ROC 曲线

ROC curve for decision trees using rocr package

r

classification

data-mining

decision-tree