Caret Library Example - R Language

caret_examples.r

# caret library example used in blogpost:
# https://towardsdatascience.com/a-guide-to-using-caret-in-r-71dec0bda208

library(caTools)
library(caret)

# Train Test Split on both Iris and Mtcars

train_test_split <- function(df) {
  set.seed(42)
  sample = sample.split(df, SplitRatio = 0.8)
  train = subset(df, sample == TRUE)
  test  = subset(df, sample == FALSE)
  return (list(train, test))
}

# To Simplify, we are going to make iris a two classifiction
# problem

iris$target <- ifelse(
  iris$Species == 'setosa',
  1,
  0
)

# mtcars
mtcars_train <- train_test_split(mtcars)[[1]]
mtcars_test <- train_test_split(mtcars)[[2]]

# iris
iris_train <- train_test_split(iris)[[1]]
iris_test <- train_test_split(iris)[[2]]

# Training a Linear Model

lm_model <- train(mpg ~ hp + wt + gear + disp, 
      data = mtcars_train, 
      method = "lm")

summary(lm_model)

# Training a Logistic Regression

glm_model <- train(target ~ Sepal.Length + Sepal.Width + Petal.Length + Petal.Width, 
                  data = iris_train, 
                  method = "glm",
                  family = "binomial")

summary(glm_model)

# Training a decision tree
d.tree <- train(mpg ~ hp + wt + gear + disp, 
                  data = mtcars_train, 
                  method = "rpart")

library(rpart.plot)
rpart.plot(d.tree$finalModel)

# Examining our dtree
d.tree

# Training a Random Forest
r.forest <- train(mpg ~ hp + wt + gear + disp, 
                  data = mtcars_train, 
                  method = "ranger")

r.forest

# Training a XGBoost Model
xg.boost <- train(mpg ~ hp + wt + gear + disp, 
                  data = mtcars_train, 
                  method = "xgbTree")


# Training a k-nearest neighbor
knn <- train(mpg ~ hp + wt + gear + disp, 
                  data = mtcars_train, 
                  method = "knn")

# Training a neural network
neural.network <- train(mpg ~ hp + wt + gear + disp, 
             data = mtcars_train, 
             method = "neuralnet")

# Training a neural network for Classification
neural.network.class <- train(target ~ Sepal.Length + Sepal.Width + Petal.Length + Petal.Width, 
                        data = iris_train, 
                        method = "nnet")

# Results of Neural Network Class
neural.network.class

iris_train$target = factor(iris_train$target)

# Training a neural network for Classifiction with Factor
neural.network.class.2 <- train(target ~ Sepal.Length + Sepal.Width + Petal.Length + Petal.Width, 
                              data = iris_train, 
                              method = "nnet")

# K-Fold Cross Validation

ctrl<- trainControl(method="cv", number=10)

d.tree.kfold <- train(mpg ~ hp + wt + gear + disp, 
                data = mtcars_train,
                trControl = ctrl,
                method = "rpart")
d.tree.kfold

rpart.plot(d.tree.kfold$finalModel)

# Optimizing for Another Metric
d.tree.kfold.rsquared <- train(mpg ~ hp + wt + gear + disp, 
                      data = mtcars_train,
                      trControl = ctrl,
                      metric = "Rsquared",
                      method = "rpart")

d.tree.kfold.rsquared


# Hyperparameter Tuning
d.tree.hyperparam <- train(mpg ~ hp + wt + gear + disp, 
                               data = mtcars_train,
                               trControl = ctrl,
                               metric = "Rsquared",
                               method = "rpart",
                               tuneGrid = data.frame(
                                 cp = c(0.0001, 0.001, 0.35, 0.65)))

d.tree.hyperparam


r.forest.hyperparam <- train(mpg ~ hp + wt + gear + disp, 
                           data = mtcars_train,
                           trControl = ctrl,
                           metric = "Rsquared",
                           method = "ranger",
                           tuneGrid = data.frame(
                             mtry = c(2, 3, 4),
                             min.node.size = c(2, 4, 10),
                             splitrule = c('variance')))

r.forest.hyperparam

# Predicting On the Test Set

# Linear Regression
predict(lm_model, mtcars_test)

# Decision Tree 
predict(d.tree, mtcars_test)

# Random Forest
predict(r.forest, mtcars_test)

# XGBoost
predict(xg.boost, mtcars_test)

# Comparing Predictions

predictions <- data.frame(
  rf = predict(r.forest, mtcars_test),
  xgboost = predict(xg.boost, mtcars_test)
)

library(ggplot2)

ggplot(
  data = predictions,
  aes(x = rf, y = xgboost)
) + geom_point()