从多个 confusionMatrix 中提取指标并填充数据框

Question

我有许多用于各种不同分类模型的 confusionMatrix 输出。我想从每个混淆矩阵中提取模型指标（即灵敏度 --> 平衡精度，或来自 cf_1$byClass 的所有指标）来填充数据框。我希望数据框包含所有这些指标，以及它来自哪个混淆矩阵的标签。谁能帮我弄清楚该怎么做？请参阅下面的可重现示例：

library(caret)
library(tidyverse)

df_1 <- data.frame(x = sample(LETTERS[1:2], 20, replace = T),
                      y = sample(LETTERS[1:2], 20, replace = T))

df_2 <- data.frame(x = sample(LETTERS[1:2], 20, replace = T),
                      y = sample(LETTERS[1:2], 20, replace = T))

cf_1 <- confusionMatrix(df_1$y, df_1$x)
cf_2 <- confusionMatrix(df_2$y, df_2$x)

# I would like a dataframe with each of these, as well as column for the associated confusion matrix name
cf_1$byClass
cf_2$byClass

除了所有相关指标外，我想要的输出如下所示：

x = c(0.2, 0.4)
y = c(0.5, 0.6)
z = c("cf_1", "cf_2"
      )

output <- data.frame(Sensitivity = x, Specificity = y, Model = z)

更新

这是我想出的解决方案，它不是很漂亮，但很有效。话虽这么说，仍然愿意看看是否有人能想出一种更清洁、更有效的方法来做到这一点！

model_names <- c("cf_1", "cf_2")
model_metrics <- list()
for (i in 1:length(model_names)) {
     for (j in model_names) {
          model <- eval(as.name(j))
          results <- model$byClass %>% round(digits = 3)
          results <- c(model = j, results)
          model_metrics[[j]] <- results
     }
}

model_metrics <- do.call(rbind, model_metrics) %>% 
     data.frame %>% 
     `rownames<-`(NULL)

model_metrics

Answer 1

怎么样：

library(caret)
library(tidyverse)

df_1 <- data.frame(x = sample(LETTERS[1:2], 20, replace = T),
                      y = sample(LETTERS[1:2], 20, replace = T))

df_2 <- data.frame(x = sample(LETTERS[1:2], 20, replace = T),
                      y = sample(LETTERS[1:2], 20, replace = T))

cf_1 <- confusionMatrix(df_1$y, df_1$x, mode = "sens_spec")
cf_2 <- confusionMatrix(df_2$y, df_2$x, mode = "sens_spec")

bind_rows(stack(cf_1$byClass), stack(cf_2$byClass), .id = "id") %>% 
  filter(ind %in% c("Sensitivity", "Specificity")) %>% 
  spread(id, values) %>% 
  rename(metric = ind, mod1 = "1", mod2 = "2")


#>        metric      mod1 mod2
#> 1 Sensitivity 0.4285714  0.5
#> 2 Specificity 0.6153846  0.4

编辑要获取数据框，只需堆叠 cf_1 列表：

data.frame(stack(cf_1$byClass))

Answer 2

两种方法（顺便说一下，我认为您的示例中有一个类型，df_1 使用了两次）

# Base R
output <- data.frame(rbind(cf_1$byClass, cf_2$byClass))
output$model = c('cf_1', 'cf_2')

# dplyr version
output <- bind_rows(cf_1$byClass, cf_2$byClass, .id = 'model')

从多个 confusionMatrix 中提取指标并填充数据框

Extract metrics and populate data frame from multiple confusionMatrix's

r

machine-learning

confusion-matrix