Compute average confusion matrix across resamples

Source: R/conf_mat_resampled.R

For classification problems, conf_mat_resampled() computes a separate confusion matrix for each resample then averages the cell counts.

conf_mat_resampled(x, parameters = NULL, tidy = TRUE)

Arguments

x

An object with class tune_results that was used with a classification model that was run with control_*(save_pred = TRUE).
parameters

A tibble with a single tuning parameter combination. Only one tuning parameter combination (if any were used) is allowed here.
tidy

Should the results come back in a tibble (TRUE) or a matrix.

Value

A tibble or matrix with the average cell count across resamples.

Examples

library(parsnip)
library(rsample)
library(dplyr)

data(two_class_dat, package = "modeldata")

set.seed(2393)
res <-
  logistic_reg() %>%
  set_engine("glm") %>%
  fit_resamples(Class ~ ., resamples = vfold_cv(two_class_dat, v = 3),
                control = control_resamples(save_pred = TRUE))

conf_mat_resampled(res)

#> # A tibble: 4 x 3
#>   Prediction Truth   Freq
#>   <fct>      <fct>  <dbl>
#> 1 Class1     Class1 124  
#> 2 Class1     Class2  25.7
#> 3 Class2     Class1  21.7
#> 4 Class2     Class2  92.3
conf_mat_resampled(res, tidy = FALSE)

#>           Class1   Class2
#> Class1 124.00000 21.66667
#> Class2  25.66667 92.33333