Assignment2.Rmd

---
title: "assignment2"
author: "vneilley"
date: "3/29/2021"
output: html_document
---

```
if(!require(rmarkdown)) install.packages("rmarkdown") 
if(!require(ISLR)) install.packages("ISLR")
if(!require(class)) install.packages("class")
if(!require(dplyr)) install.packages("dplyr")
if(!require(caret)) install.packages("caret")
if(!require(ggplot)) install.packages("ggplot2")
if(!require(sqldf)) install.packages("sqldf")
if(!require(tidyverse)) install.packages ("tidyverse")
if(!require(devtools)) install.packages ("devtools")
if(!require(rpart)) install.packages ("rpart")
if(!require(mlbench)) install.packages ("mlbench")
if(!require(rpart.plot)) install.packages ("rpart.plot")
if(!require(SuperLearner)) install.packages ("SuperLearner")
if(!require(ranger)) install.packages ("ranger")
library(ranger)
library(SuperLearner)
library(rpart)
library(mlbench)
library(skimr)
library(readr)
library(dplyr)
library(tree)
library(rpart.plot)
```

### Import Data

```
R_assignment_2_data <- read_csv("./HIA218-2/218-2.csv")
View(R_assignment_2_data)
diabetes_data <- R_assignment_2_data
diabetes_manipulated <- R_assignment_2_data
```

## Question 1 

### Do we need to do extra data preprocessing?

```
skim(diabetes_data)
```

Yes everything is characters, change to 0 and 1

```
diabetes_manipulated <- diabetes_data %>%
      mutate(Gender = ifelse(Gender == "Male",0,1),
             Polyuria = ifelse(Polyuria == "No" ,0,1),
             Polydipsia = ifelse(Polydipsia == "No" ,0,1),
             `sudden weight loss` = ifelse(`sudden weight loss` == "No" ,0,1),
             weakness = ifelse(weakness == "No" ,0,1),
             Polyphagia = ifelse(Polyphagia == "No" ,0,1),
             `Genital thrush` = ifelse(`Genital thrush` == "No" ,0,1),
             `visual blurring` = ifelse(`visual blurring` == "No" ,0,1),
             Itching = ifelse(Itching == "No" ,0,1),
             Irritability = ifelse(Irritability == "No" ,0,1),
             `delayed healing` = ifelse(`delayed healing` == "No" ,0,1),
             `partial paresis` = ifelse(`partial paresis` == "No" ,0,1),
             `muscle stiffness` = ifelse(`muscle stiffness` == "No" ,0,1),
             Alopecia = ifelse(Alopecia == "No" ,0,1),
             Obesity = ifelse(Obesity == "No" ,0,1),
             class = ifelse(class == "Negative" ,0,1))
```

### Are there missing values?

No.

### Outliers

```
boxplot(diabetes_manipulated$Age, plot=TRUE)

###Outliers above 80 deleted.

diabetes_manipulated <- diabetes_manipulated[!diabetes_manipulated$Age>80,]
```

The rest of data is binary so will not be removing outliars.

## Question 2 Split 70/30 for training

```
train_data <- diabetes_manipulated %>% mutate(ref=seq_len(nrow(diabetes_manipulated))) %>% sample_frac(0.7)
test_data <- diabetes_manipulated[-train_data$ref,]
```

## Question 3 Build the tree

```
diabetes_tree <- rpart(class~.,
                       data = diabetes_manipulated,
                       method = "class")
```

```
diabetes_tree
rpart.plot(diabetes_tree)
```

### Question 4 Build ensemble method

```
model = glm(class~., family = "binomial", data = diabetes_manipulated)
```

```
predict = predict(model, newdata = test_data, type = "response")
```

```
table(test_data$class, predict >= 0.5)
```