R Trees

Question 1

Create Categorical Variable From Contious Var

Answer 1

High = ifelse(Sales)

Question 2

Add Column to data frame

Answer 2

Carseats = data.frame(Carseats, High) #Add column to data frame

Question 3

Fitting tree model

Answer 3

If doing it to categorical variable
tree.carseats = tree(High~.-Sales, Carseats) # Want to fit the model to all variables besides Sales, which we converted to categorical earlier
summary(tree.carseats)

Question 4

Pruning Decision Trees

Answer 4

cv. carseats=cv.tree(tree.carseats, FUN = prune.misclass)
cv. carseats

pick the tree size that has the lowest dev

Question 5

library for decision trees

Answer 5

library(tree)

Question 6

library for regression trees

Answer 6

library(tree) is the same thing as a decision tree except the outcome is a continous variable instead of 0/1

Question 7

library for random forest

Answer 7

library(randomForest)

Question 8

how to get variable importance in random forest

Answer 8

importance(modeloutput)

Question 9

library and function for random forest with boosting

Answer 9

library(gbm)

boosted.forest = gbm(y~., data = DF, distribution…etc)

Question 10

what are the tuning parametrs for a boosted regression or decision tree?

Answer 10

#Parameters to Boosting Model - Really 3 Tuning Parameters
# 1) The Number of Trees (n.trees), we can overfit if this is too large, want to choose this with CV
# 2) Distribution = 'gaussian' if regression model, 'bernoulli' if classification model
# 3) shrinkage parameter = you can usually leave this at default, but can tweak this using CV, default is .001
# 4) Interaction Depth (d), d = 1 usually works well, but idea is to keep this small can also select this using CV