treemisc (version 0.0.1)

decision_boundary: Add decision boundary to a scatterplot.

Description

Adds the decision boundary from a classification model (binary or multiclass) to an existing scatterplot.

Usage

decision_boundary(model, train, y, x1, x2, pfun, grid.resolution = 100, ...)
# S3 method for default
decision_boundary(
  model,
  train,
  y,
  x1,
  x2,
  pfun = NULL,
  grid.resolution = 100,
  ...
)

Value

No return value, only called for side effects; in this case, a contour displaying the decision boundary of a classifier is added to an existing scatterplot.

Arguments

model

The associated model object.

train

Data frame of training observations.

y

Character string giving the name of the outcome variable in train.

x1

Character string giving the name of the predictor in train that corresponds to the x-axis.

x2

Character string giving the name of the predictor in train that corresponds to the y-axis.

pfun

Optional prediction wrapper that returns a vector of predicted class labels. It must have exactly two arguments: object and newdata.

grid.resolution

Integer specifying the resolution of the contour plot. Default is 100.

...

Additional optional arguments to be passed on to contour.

Examples

if (FALSE) {
library(mlbench)
library(rpart)
library(treemisc)

# Generate training data from the twonorm benchmark problem
set.seed(1050)  # for reproducibility
trn <- as.data.frame(mlbench.twonorm(500, d = 2))

# Fit a default classification tree
tree <- rpart(classes ~ ., data = trn)

# Scatterplot of training data
palette("Okabe-Ito")
plot(x.2 ~ x.1, data = trn, col = as.integer(trn$classes) + 1,
     xlab = expression(x[1]), ylab = expression(x[2]))
palette("default")

# Add a decision boundary
decision_boundary(tree, train = trn, y = "y", x1 = "x.1", x2 = "x.2")
}