DataCamp Supervised
Learning in R: Classification
SUPERVISED LEARNING IN R: CLASSIFICATION
Making decisions with
trees
Brett Lantz
Instructor
DataCamp Supervised Learning in R: Classification
A decision tree model
DataCamp Supervised Learning in R: Classification
Decision trees for prediction
DataCamp Supervised Learning in R: Classification
Divide-and-conquer
DataCamp Supervised Learning in R: Classification
Divide-and-conquer
DataCamp Supervised Learning in R: Classification
Divide-and-conquer
DataCamp Supervised Learning in R: Classification
The resulting tree
DataCamp Supervised Learning in R: Classification
Building trees in R
# building a simple rpart classification tree
library(rpart)
m <- rpart(outcome ~ loan_amount + credit_score, data = loans,
method = "class")
# making predictions from an rpart tree
p <- predict(m, test_data, type = "class")
DataCamp Supervised Learning in R: Classification
SUPERVISED LEARNING IN R: CLASSIFICATION
Let's practice!
DataCamp Supervised Learning in R: Classification
SUPERVISED LEARNING IN R: CLASSIFICATION
Growing larger
classification trees
Brett Lantz
Instructor
DataCamp Supervised Learning in R: Classification
Choosing where to split
DataCamp Supervised Learning in R: Classification
Axis-parallel splits
DataCamp Supervised Learning in R: Classification
The problem of overfitting
DataCamp Supervised Learning in R: Classification
Evaluating model performance
DataCamp Supervised Learning in R: Classification
SUPERVISED LEARNING IN R: CLASSIFICATION
Let's practice!
DataCamp Supervised Learning in R: Classification
SUPERVISED LEARNING IN R: CLASSIFICATION
Tending to
classification trees
Brett Lantz
Instructor
DataCamp Supervised Learning in R: Classification
Pre-pruning
DataCamp Supervised Learning in R: Classification
Post-pruning
DataCamp Supervised Learning in R: Classification
Pre- and post-pruning with R
# pre-pruning with rpart
library(rpart)
prune_control <- [Link](maxdepth = 30, minsplit = 20)
m <- rpart(repaid ~ credit_score + request_amt,
data = loans,
method = "class",
control = prune_control)
# post-pruning with rpart
m <- rpart(repaid ~ credit_score + request_amt,
data = loans,
method = "class")
plotcp(m)
m_pruned <- prune(m, cp = 0.20)
DataCamp Supervised Learning in R: Classification
SUPERVISED LEARNING IN R: CLASSIFICATION
Let's practice!
DataCamp Supervised Learning in R: Classification
SUPERVISED LEARNING IN R: CLASSIFICATION
Seeing the forest from
the trees
Brett Lantz
Instructor
DataCamp Supervised Learning in R: Classification
Understanding random forests
DataCamp Supervised Learning in R: Classification
Making decisions as an ensemble
DataCamp Supervised Learning in R: Classification
Random forests in R
# building a simple random forest
library(randomForest)
m <- randomForest(repaid ~ credit_score + request_amt, data = loans,
ntree = 500, # number of trees in the forest
mtry = sqrt(p)) # number of predictors (p) per tree
# making predictions from a random forest
p <- predict(m, test_data)
DataCamp Supervised Learning in R: Classification
SUPERVISED LEARNING IN R: CLASSIFICATION
Let's practice!