Decision Trees

Introduction

Decision trees split data based on feature values to make predictions. They're interpretable and useful for classification and regression.

Building Trees

library(rpart)

# Classification tree
tree <- rpart(target ~ predictors, data = train)

# Print tree
print(tree)

# Plot tree
plot(tree)
text(tree)

Predictions

# Class predictions
predict(tree, test, type = "class")

# Probability predictions
predict(tree, test, type = "prob")

Pruning

# Find optimal complexity
plotcp(tree)

# Prune tree
pruned_tree <- prune(tree, cp = 0.05)

Using Caret

library(caret)
train(target ~ ., data = train, method = "rpart")

Summary

Decision trees are interpretable. Use pruning to avoid overfitting.