CODEWITHSUNDEEP
rplotrandom-forestdecision-tree
ECII
ECII

Reputation: 10629

How to plot a decision boundary of random forest model

I have 

## Classification:
library("randomForest")
data=iris
data<-data[data$Species!="setosa",]
data$Species<-factor(as.character(data$Species))
iris.rf <- randomForest(Species ~ Sepal.Length+Petal.Length, data=data, importance=TRUE,            proximity=TRUE)

I would like to construct a Sepal.Length~Petal.Length with the decision boundary. And what kind of boundary will this be? 0.5 probability for each of the 2 classes?

Upvotes: 1

Views: 2624

Answers (1)

C8H10N4O2
C8H10N4O2

Reputation: 19005

You have a random forest, so there is not necessarily a clear decision boundary like you would get from a non-probabilistic linear classifier like SVM. But you could plot it with something like...

library(ggplot2)
ggplot(data=data,aes(x=Petal.Length, y=Sepal.Length, color= iris.rf$predicted) ) + 
       geom_point()

enter image description here

And in this case yes, since you trained it only on two classes, the boundary represented by the color change happens at 0.5.

Upvotes: 1

Related Questions