IS - Skripta 4

##########################
# Classification trees
##########################

# load ISLR package
# install.packages('ISLR')

# get the Carseats dataset docs

# examine dataset structure

# examine Sales variable distribution

# get the 3rd quartile of the Sales variable

# create a new variable HighSales based on the value of the Sales variable

# check the type of the HighSales variable

# convert HighSales into a factor variable

# get the distribution of the HighSales variable

##################################
# Create train and test datasets
##################################

# remove Sales variable

# load caret package

# create train and test datasets

# print distributions of the outcome variable on train and test datasets

#######################################################
# Create a prediction model using Classification Trees
#######################################################

# load rpart library

# rpart uses random sampling, so, we have to set the seed value before calling the function

# build the model

# print the model

# load rpart.plot library
# install.packages("rpart.plot")

# plot the tree


# make the predictions with tree1 over the test dataset

# print several predictions

# create the confusion matrix

# function for computing evaluation measures

# compute the evaluation metrics

# get the docs for the rpart.control function

# build the second model with minsplit = 10 and cp = 0.001

# print the model

# plot the tree2

# make the predictions with tree2 over the test dataset

# create the confusion matrix for tree2 predictions

# compute the evaluation metrics

# compare the evaluation metrics for tree1 and tree2

# load e1071 library
# install.packages('e1071')

# define cross-validation (cv) parameters; we'll do 10-fold cross-validation

# then, define the range of the cp values to examine in the cross-validation

# since cross-validation is a probabilistic process, we need to set the seed
# so that the results can be replicated

# run the cross-validation

# plot the cross-validation results


# create tree2 using the new cp value

# plot the new tree

# make the predictions with tree3 over the test dataset

# create the confusion matrix for tree3 predictions

# compute the evaluation metrics

# compare the evaluation metrics for tree1, tree2 and tree3