plotClassification2D: Visualize a classification task in 2D

In BodoBurger/bodomisc: Collection of helper functions mainly for interactive use of R

Description

Plot decision boundaries, prediciton areas and original data for two features.

Usage

plotClassification2D(model, task, features, grid.res = 100,
  x1.lim = NULL, x2.lim = NULL, colours = FALSE)

Arguments

model	[mlr WrappedModdel object]
task	[mlr classification task object] Created by mlr::makeClassifTask
features	[character(2)] Names of two numeric features.
grid.res	[numeric(1)]
x1.lim	[numeric(2)]
x2.lim	[numeric(2)]
colours	[character(1)] "ESL": colours from the book "Elements of statistical learning"

Value

ggplot object

Examples

library(mlr)
library(ggplot2)
theme_set(theme_light())

# visualize randomForest predictions on iris data
require(randomForest)
iris.mod.rf = train(makeLearner("classif.randomForest"), iris.task)
plotClassification2D(iris.mod.rf, iris.task, features = c("Petal.Length", "Petal.Width"))

# recreate plots from chapter 2 of "Elements of Statistical Learning"
require(ElemStatLearn)
me = ElemStatLearn::mixture.example
df = data.frame(x1 = me$x[,1], x2 = me$x[,2], y = factor(me$y))
tsk = makeClassifTask(data = df, target = "y")
spam.knn = train(makeLearner("classif.knn", k = 15), tsk)
plotClassification2D(spam.knn, tsk, features = c("x1", "x2"), colours = "ESL")

BodoBurger/bodomisc documentation built on May 27, 2020, 5:12 p.m.