Homework 3.Rmd

---
title: "Homework 3"
author: "Jiawen Qi (jiq10)"
date: "October 20, 2016"
output: word_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

## 1. Read in the iris data

Iris data is embedded in Rstudio. We can directly use this one.

```{r}
iris <- iris
summary(iris) # get a summary
head(iris) # check the data
```

## 2. We only use sepal length as a variable

```{r}
iris_1 <- iris[,-c(2, 3, 4)]
summary(iris_1) # get a summary
head(iris_1) # check the data
```

## 3. Create a 10 fold knn execution with k = 5.And from the testing set keep track of what each observation was predicted to be.

```{r}
iris_1$random <- runif(nrow(iris_1)) # add one random column
iris_1 <- iris_1[order(iris_1$random),] # order by random number
for (i in 1:nrow(iris)) {
  iris_1$group[i] <- (i%%10) + 1 # make 10 groups
}
iris_1$results <- c(rep("predicted",150)) # create a new column to store the predict result
table(iris_1$Species,iris_1$group) # these groups are random
library(class) # open the library for knn
for (i in 1:10) {
  train.data <- as.matrix((iris_1[iris_1$group != i,1])) # get the value of sepal length
  train.class <- as.matrix((iris_1[iris_1$group != i,2])) # get the class
  test.data <- as.matrix((iris_1[iris_1$group == i,1])) # get the date to be predicted
  knn5 = knn(train.data,            # training data as a matrix
             test.data,             # testing data as a matrix
             train.class,           # class for training data
             k=5)                   # k = 5
  test.predicted <- data.frame(test.data, pred = knn5) # get predicted values for test
  iris_1$results[iris_1$group == i] <- test.predicted$pred  # merge predicted into iris
}
```

## 4. Create a confusion matrix of the final results.

```{r}
confusion_matrix <- table(iris_1$results, iris_1$Species) # the results matrix
print(confusion_matrix) # print the confusion matrix
row.names(confusion_matrix) <- c("setosa", "versicolor", "virginica") # change the row name and print again
print(confusion_matrix)
```

## 5. Create a density plot of correct and incorrect predictions as a function of sepal length.

```{r}
iris_1$Species <- as.character(iris_1$Species)
iris_1$results <- as.character(iris_1$results)
iris_1$result.lable <- "FALSE" 
for (i in 1:150) {
  if(iris_1$Species[i] %in% c("setosa") && iris_1$results[i] %in% c("1")) {
  iris_1$result.lable[i] <- "TRUE" 
  }
  if(iris_1$Species[i] %in% c("versicolor") && iris_1$results[i] %in% c("2")) {
  iris_1$result.lable[i] <- "TRUE" 
  }
  if(iris_1$Species[i] %in% c("virginica") && iris_1$results[i] %in% c("3")) {
  iris_1$result.lable[i] <- "TRUE" 
  }
}
iris_1$result.lable <- as.factor(iris_1$result.lable)
cdplot(result.lable ~ Sepal.Length, 
       data = iris_1, 
       main = "density plot of predict results ") # conditional density plot
```