Decision_Tree.Rmd

---
title: "Decision Tree Try Out"
author: "Melvin Seinen"
date: "1/4/2022"
output: word_document
---

```{r}

install.packages('caret')
install.packages('rpart')
install.packages('rpart.plot')
install.packages('ROCR')
# Load the packages.
library(caret) # Functions to streamline model training and tuning processes
library(rpart) # Implementation of CART analysis algorithm
library(rpart.plot) # Procedure to plot the results of rpart
library(ROCR)


load('eclsk.rdata')

#Kindergarten measures of specific math knowledge (CountNumShp, RelativeSize,
#OrdinalSeq, and AddSubtract)
#Kindergarten measures of language knowledge (LetterRecog, BeginSounds,
#EndSounds, and SightWords)
#Kindergarten measures of broad knowledge (LiteracyARS, MathARS, and
#GeneralARS).
#The CART analysis uses these independent variables to predict the indicator for scoring
#Below Proficient on the state math assessment at the end of grade 3 (AtRisk),


# Use the load() command to read in the data.
load('eclsk.rdata') # Change the file name to access your data file.
# Create a copy of your data in mydata to allow you to use the remaining code more easily.
mydata <- eclsk # Change the dataset name to access your data.
# Assign the outcome to depvar, the dependent variable.
depvar <- 'AtRisk' # Change the dependent variable to your outcome of interest.
# Define the set of independent variables for the analysis.
indepvar <- c( # Change the independent variables to your characteristics.
'LiteracyARS','MathARS','GeneralARS','LetterRecog','BeginSounds',
'EndSounds','SightWords','CountNumShp','RelativeSize','OrdinalSeq',
'AddSubtract')
# Use dim() to show the dimensions of the data.
dim(mydata)

set.seed(101010)


train_index <- as.vector(createDataPartition(mydata[[depvar]],p=.8,list=FALSE,times=1))


mytrain <- mydata[train_index, ]
mytest <- mydata[-train_index, ]


myformula <- as.formula(paste(depvar,paste(indepvar,collapse=' + '),sep=' ~ '))
# Use trainControl() to define parameters for train().
mycontrol <- trainControl(
method = 'repeatedcv', # Repeated cross-validation
number = 10, # Number of folds (k)
repeats = 10, # Number of repeats (n) of cross-validation
savePredictions = 'final', # Save predictions for best tuning parameters
classProbs = TRUE, # Compute probabilities for each class
selectionFunction = 'oneSE', # Select model within one standard error of best
summaryFunction = twoClassSummary # Provide ROC, sensitivity, and specificity
)

myformula
mycontrol
# Use caret’s train() function to tune using cp and select using ROC.
mytree <- train(
myformula, # Use the formula defined above
data = mytrain, # Use the subset of data for training
method = 'rpart', # Use the rpart procedure for CART analysis
trControl = mycontrol, # Use the controls defined above
tuneLength = 10, # Try 10 values of the complexity parameter (cp)
metric='ROC'
)

mytree

mytree$finalModel


rpart.plot(
mytree$finalModel, # The optimal model from the CART analysis
box.palette = 'lightblue1', # Box color
type = 0, # Draw labels for each split and node
leaf.round = 0, # Do not use rounded terminal nodes
nn = T, # Include node numbers
branch.col = 'lightblue2', # Color of the branches
branch.type = 5, # Branch width based on share of students
extra = 107, # Show % in node at risk and % of all students
xflip = T, # Flip the tree horizontally
under = T, # Place overall percentage under leaf
cex = 1 # Size of text
)

```