3_CART_iris.Rmd

---
title: "CART (iris)"
output: pdf_document
---

1. Use the random seed 123 to divide the cleaned data into 80% training and 20% testing. 
```{r}
if (!requireNamespace("tidyverse")) install.packages('tidyverse')
if (!requireNamespace("caret")) install.packages('caret')
if (!requireNamespace("rpart")) install.packages('rpart')
if (!requireNamespace("rpart.plot")) install.packages('rpart.plot')
if (!requireNamespace("rattle")) install.packages('rattle')
if (!requireNamespace("nnet")) install.packages('nnet')
library(tidyverse)
library(caret)
library(rpart)
library(rpart.plot)
library(rattle)
library(nnet)

data('iris')
data = iris
str(data)

set.seed(123)
training.samples <- data$Species %>% 
  createDataPartition(p = 0.8, list = FALSE)
train.data  <- data[training.samples, ]
test.data <- data[-training.samples, ]
```

2. Fully grown tree
```{r}
# Basic tree with default parameters
model <- rpart(Species ~., data = train.data, cp = 0)
# The default splitting method for Speciesification is 'gini'
# To define the split criteria, you use parms=list(split='...')
par(xpd = NA)
fancyRpartPlot(model) # rattle pkg
rpart.plot(model) # another way to print the decision tree
pred1 <- predict(model,newdata = test.data, type ='class')
confusionMatrix(factor(pred1), factor(test.data$Species))
```

3 & 4. Prune the fully grown tree using the training set with 10_fold CV
```{r}
set.seed(123)
model2 <- train(
  Species ~., data = train.data, method = "rpart",
  trControl = trainControl("cv", number = 10),
  tuneLength = 100)
plot(model2)
model2$bestTune
fancyRpartPlot(model2$finalModel)

pred2 <- predict(model2, newdata = test.data)
confusionMatrix(factor(pred2), factor(test.data$Species))
```

5. Multinomial logistic regression
```{r}
model3 <- multinom(Species ~ . , data = train.data)
coef(model3)
pred3 <- model3 %>% predict(test.data, type = "class") 
confusionMatrix(factor(pred3), factor(test.data$Species))
```