Data Science Tutorials Blog Post Series: Training a Decision Tree using R

decisiontree.R

# Training a decision tree in R - used in blog post:
# https://medium.com/codex/data-science-tutorials-training-a-decision-tree-using-r-d6266936d86

library(dplyr)
library(rpart)
library(rpart.plot)
library(caret)
library(Metrics)
library(ggplot2)

# Load London Bike CSV File
london_bike <- read.csv('./london_merged.csv')

# Using a function to split into train and test
train_test_split <- function(data, percentage) {
  
  data_with_row_id <- data %>% 
    mutate(id = row_number())
  
  set.seed(1234)
  training_data <- data_with_row_id %>%
    sample_frac(percentage)
  test_data <- anti_join(
    data_with_row_id,
    training_data,
    by='id'
  )
  
  training_data$id <- NULL
  test_data$id <- NULL
  
  return (list(training_data, test_data))
}

# Keeping 80% for the training set
training_data <- train_test_split(
  london_bike, 0.8
  )[[1]]
test_data <- train_test_split(
  london_bike, 0.8
  )[[2]]

# Subsetting only the features and target
training_data <- training_data[,c('t1','t2','hum',
                            'wind_speed','weather_code',
                            'is_holiday','is_weekend',
                            'season', 'cnt')]

test_data <- test_data[,c('t1','t2','hum',
                       'wind_speed','weather_code',
                       'is_holiday','is_weekend',
                       'season', 'cnt')]


# Fitting Decision Tree Model using rpart
set.seed(1234)
dtree_1 <- rpart(formula = cnt ~ .,
                 data = training_data,
                 control = list(minbucket=120, maxdepth=4),
                 method = 'anova'
          )

# Plotting the Tree
prp(dtree_1)

# Timing the execution
system.time(
  dtree_1 <- rpart(formula = cnt ~ .,
                   data = training_data,
                   control = list(minbucket=120, maxdepth=4),
                   method = 'anova'
  ))

# Assessing performance using RMSE
rmse(
  test_data$cnt, 
  predict(dtree_1, test_data)
  )

# Adding a new hyperparameter
set.seed(1234)
dtree_2 <- rpart(formula = cnt ~ .,
                 data = training_data,
                 control = list(minbucket=10, 
                                maxdepth=10, 
                                minsplit=10,
                                cp=0.001),
                 method = 'anova'
)

# Checking parameters using help function
?rpart.control

# Assessing execution time
system.time(rpart(formula = cnt ~ .,
                  data = training_data,
                  control = list(minbucket=10, 
                                 maxdepth=10, 
                                 minsplit=10,
                                 cp=0.001),
                  method = 'anova'))


# Assessing performance
rmse(
  test_data$cnt, 
  predict(dtree_2, test_data)
)

# Plotting the extended tree
prp(dtree_2)

# Using Caret
caret_tree <- train(form = cnt ~ .,
      data = training_data,
      method="rpart", 
      control = rpart.control(minsplit = 10, 
                              minbucket = 10,
                              maxdepth = 10
                              ))

# Assessing execution time
system.time(train(form = cnt ~ .,
      data = training_data,
      method="rpart", 
      control = rpart.control(minsplit = 10, 
                              minbucket = 10,
                              maxdepth = 10
      )))

# Drawing our Tree trained with Caret
prp(caret_tree$finalModel)

# Assessing performance of Caret Tree
rmse(
  test_data$cnt, 
  predict(caret_tree$finalModel, test_data)
)