Draw Probability regions for Classification model

This function is mostly useful in an educational setting. Can only be used with trained workflow objects with 2 numeric predictor variables.

viz_prob_region(x, new_data, resolution = 100, expand = 0.1, facet = FALSE)

Arguments

x	trained `workflows::workflow` object.
new_data	A data frame or tibble for whom the preprocessing will be applied.
resolution	Number of squared in grid. Defaults to 100.
expand	Expansion rate. Defaults to 0.1. This means that the width and height of the shaded area is 10 data.
facet	Logical, whether to facet chart by class. Defaults to FALSE. The chart have been minimally modified to allow for easier styling.

Value

`ggplot2::ggplot` object

Examples

library(parsnip)
library(workflows)

iris2 <- iris
iris2$Species <- factor(iris2$Species == "setosa",
                        labels = c("setosa", "not setosa"))

svm_spec <- svm_rbf() %>%
  set_mode("classification") %>%
  set_engine("kernlab")

svm_fit <- workflow() %>%
  add_formula(Species ~ Petal.Length + Petal.Width) %>%
  add_model(svm_spec) %>%
  fit(iris2)

viz_prob_region(svm_fit, iris2)

viz_prob_region(svm_fit, iris2, resolution = 20)

viz_prob_region(svm_fit, iris2, expand = 1)

viz_prob_region(svm_fit, iris2, facet = TRUE)

knn_spec <- nearest_neighbor() %>%
  set_mode("classification") %>%
  set_engine("kknn")

knn_fit <- workflow() %>%
  add_formula(class ~ umap_1 + umap_2) %>%
  add_model(knn_spec) %>%
  fit(mnist_sample)

viz_prob_region(knn_fit, mnist_sample, facet = TRUE)