在R(插入符号)中训练模型的pred值的 NA

Question

set.seed(2)
rpcv <- trainControl(method='repeatedcv', number=4, repeats = 10,
savePredictions = T, classProbs = T)
iris2 <- iris[c(1:3,60:72,100:109),]
iris2_train <- iris2[-1,]
iris2_test <- iris2[1,]
set.seed(4)
iris_svm <- train(as.factor(Species)~., data=iris2_train, method='svmRadial', trControl=rpcv)
iris_svm$pred

如果你看一下iris$pred，你可以看到有一个NA值。有什么问题吗?

Answer 1

我认为你的训练数据集有少量的样本类集(只有2个样本)，这太小了，所以运行具有足够大的n和类平衡的模型

so try this

library(caret)
set.seed(2)
rpcv <- trainControl(method='repeatedcv', number=4, repeats = 10,
savePredictions = T, classProbs = T)
# here i increased the sample of class setosa
iris2 <- iris[c(1:10,60:72,100:109),]
iris2_train <- iris2[-1,]
iris2_test <- iris2[1,]
set.seed(4)
iris_svm <- train(as.factor(Species)~., data=iris2_train, method='svmRadial', trControl=rpcv)
iris_svm$pred