patient4.Rmd

---
title: "Patient4"
author: "Anuhya B S"
date: '2022-06-08'
output:
  pdf_document: default
  word_document: default
---

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

```{r}
#Importing libraries
library('R.matlab')
library(caTools)
library(e1071)
library(class)
library(tree)
library(randomForest)
```

```{r}
#Loading Data
p1 <- readMat("data-science-P4.mat")

info <- as.data.frame(p1[2])
info <- t(info)
info <- as.data.frame(info)
lab.grp <-as.data.frame(matrix(nrow=0,ncol=1))
lab.wrd <-as.data.frame(matrix(nrow=0,ncol=1))
for (i in 1:360){
  lab.grp <-  rbind(lab.grp,info$cond[[i]])
  lab.wrd <-  rbind(lab.wrd,info$word[[i]])
}

p1.data <- p1$data
voxels <-as.data.frame(matrix(nrow=0,ncol=21764))
for (i in 1:360){
  voxels <-  rbind(voxels,p1.data[[i]][[1]])
}

```

```{r}
# Principal Component Analysis for Feature Reduction
pr.out <- prcomp(voxels)
cumsum((pr.out$sdev^2)/sum(pr.out$sdev^2))
pcs <- as.data.frame(pr.out$x[,1:300])
pcs$grp <- lab.grp$V1
#pcs$wrd <- lab.wrd$V1
```

```{r}
# Splitting data into training and test data
set.seed(100) 
#sample <- sample(1:nrow(pcs), 300)
pcs.train <- pcs[1:300,]
pcs.test <- pcs[301:360,]
#pcs.train <- subset(pcs, sample == TRUE)
#pcs.test  <- subset(pcs, sample == FALSE)
pcs.train.x <- subset(pcs.train, select = -c(grp))
pcs.train.labs <- pcs.train$grp
pcs.test.x <- subset(pcs.test, select = -c(grp))
pcs.test.labs <- pcs.test$grp
```

```{r}
# Classification Algorithms

# Naive Bayes Classifier

nb.fit <- naiveBayes(grp ~ . , data = pcs.train)
nb.class <- predict(nb.fit,pcs.test.x)
nb.class

confusion_mat.nb = as.matrix(table(Actual_Values = pcs.test.labs, Predicted_Values = nb.class)) 
print(confusion_mat.nb)

print(mean(nb.class == pcs.test$grp))
```
```{r}
# KNN

knn.pred <- knn(pcs.train.x, pcs.test.x, pcs.train.labs, k=5)

confusion_mat.knn = as.matrix(table(pcs.test.labs, knn.pred)) 
print(confusion_mat.knn)

print(mean(knn.pred == pcs.test$grp))
```

```{r}
# Decision Trees
set.seed(100)
tree.fit <- tree(as.factor(grp) ~ ., data = pcs.train)
summary(tree.fit)
plot(tree.fit)
text(tree.fit, pretty = 0)

tree.pred <- predict(tree.fit, newdata = pcs.test, type = "class")
tree.pred
confusion_mat.dt = as.matrix(table(pcs.test.labs, tree.pred)) 
print(confusion_mat.dt)

print(mean(tree.pred == pcs.test$grp))
```

```{r}
# Random Forest

rf.fit <- randomForest(as.factor(grp) ~ ., data = pcs.train,, mtry = 80, importance = TRUE)
summary(rf.fit)
rf.pred <- predict(rf.fit, newdata = pcs.test, type = "class")
confusion_mat.rf = as.matrix(table(pcs.test.labs, rf.pred)) 
print(confusion_mat.rf)

print(mean(rf.pred == pcs.test$grp))
```

```{r}
manmade <- c("furniture", "clothing", "manmade", "tool", "kitchen", "vehicle", "building", "buildpart")
natural <- c("insect", "animal", "vegetable", "bodypart")
df_new <- within(pcs, {
 cls <- "manmade"
 cls[grp %in% manmade] <- "manmade"
 cls[grp %in% natural] <- "natural"
})
pcs$cls <- df_new$cls
```

```{r}
# Splitting data into training and test data
set.seed(100) 
#sample <- sample(1:nrow(pcs), 300)
pcs.train <- pcs[1:300,]
pcs.test <- pcs[301:360,]
#pcs.train <- subset(pcs, sample == TRUE)
#pcs.test  <- subset(pcs, sample == FALSE)
pcs.train.x <- subset(pcs.train, select = -c(grp,cls))
pcs.train.labs <- pcs.train$cls
pcs.test.x <- subset(pcs.test, select = -c(grp,cls))
pcs.test.labs <- pcs.test$cls
```

```{r}
# Classification Algorithms

# Naive Bayes Classifier

nb.fit <- naiveBayes(cls ~ . , data = pcs.train)
nb.class <- predict(nb.fit,pcs.test.x)
nb.class

confusion_mat.nb = as.matrix(table(Actual_Values = pcs.test.labs, Predicted_Values = nb.class)) 
print(confusion_mat.nb)

print(mean(nb.class == pcs.test$cls))
```
```{r}
# KNN

knn.pred <- knn(pcs.train.x, pcs.test.x, pcs.train.labs, k=3)

confusion_mat.knn = as.matrix(table(pcs.test.labs, knn.pred)) 
print(confusion_mat.knn)

print(mean(knn.pred== pcs.test$cls))
```

```{r}
# Decision Trees
set.seed(100)
tree.fit <- tree(as.factor(cls) ~ ., data = pcs.train)
summary(tree.fit)
plot(tree.fit)
text(tree.fit, pretty = 0)

tree.pred <- predict(tree.fit, newdata = pcs.test, type = "class")
tree.pred
confusion_mat.dt = as.matrix(table(pcs.test.labs, tree.pred)) 
print(confusion_mat.dt)

print(mean(tree.pred== pcs.test$cls))
```

```{r}
# Random Forest

rf.fit <- randomForest(as.factor(cls) ~ ., data = pcs.train,, mtry = 80, importance = TRUE)
summary(rf.fit)
rf.pred <- predict(rf.fit, newdata = pcs.test, type = "class")
confusion_mat.rf = as.matrix(table(pcs.test.labs, rf.pred)) 
print(confusion_mat.rf)

print(mean(rf.pred== pcs.test$cls))
```