如何计算插入符号中准确度和 kappa 的 95% CI

How to calculate 95% CI for accuracy and kappa in caret

我正在使用 caret 包进行 运行 k 次重复训练,我想计算我的准确度指标的置信区间。本教程打印一个插入符号训练对象,该对象显示 accuracy/kappa 指标和相关的 SD:https://machinelearningmastery.com/tune-machine-learning-algorithms-in-r/。但是,当我这样做时,列出的都是公制平均值。

control <- trainControl(method="repeatedcv", number=10, repeats=3, search="grid")
set.seed(12345)
tunegrid <- expand.grid(.mtry=4)
rf_gridsearch <- train(as.factor(gear)~., data=mtcars, method="rf", 
                       metric="Accuracy", 
                       tuneGrid=tunegrid, 
                       trControl=control)
print(rf_gridsearch)
> print(rf_gridsearch)
Random Forest 

32 samples
10 predictors
 3 classes: '3', '4', '5' 

No pre-processing
Resampling: Cross-Validated (10 fold, repeated 3 times) 
Summary of sample sizes: 29, 28, 30, 29, 27, 28, ... 
Resampling results:

  Accuracy   Kappa    
  0.8311111  0.7021759

Tuning parameter 'mtry' was held constant at a value of 4

看起来它存储在结果对象的结果变量中。

> rf_gridsearch$results
  mtry  Accuracy     Kappa AccuracySD   KappaSD
1    4 0.7572222 0.6046465  0.2088411 0.3387574

使用 1.96 的临界 z 值可以找到 95% 的置信区间。

> rf_gridsearch$results$Accuracy+c(-1,1)*1.96*rf_gridsearch$results$AccuracySD
[1] 0.3478936 1.1665509

正确答案是:

上区间 = X_hat + z * (S/sqrt(n))

下区间 = X_hat - z * (S/sqrt(n))

如果你处理的是比例:

上区间 = X_hat + z * sqrt( (p * (1-p))/n )

下区间 = X_hat - z * sqrt( (p * (1-p))/n )