Supervised Learning -c.Rmd

---
title: "Supervised Learning Part3"
author: "Yatish"
date: "October 15, 2015"
output: html_document
---

Reading Iris data
```{r}
require(ggplot2)
require(e1071)
require(kernlab)
iris<- read.table("iris.data",sep=",")
colnames(iris)<-c("SepalLength","SepalWidth","PetalLength","petalWidth","Species")
```

## Answer 1
Classifying the data using SVM
```{r}
svm.fit<- svm(Species~.,data=iris, kernal="linear",cost=10,scale=FALSE)
plot(svm.fit,iris,SepalLength~PetalLength)
```

Classifier was able to classify iris setosa with an ease but not the other two species. Model needs improvement.

## Answer 2
Trying different kernels:
```{r}
svmfit1<-svm(Species~.,data=iris,kernal="radial",gamma=1,cost=100000)
plot(svmfit1,iris,SepalLength~PetalLength)


svm.fit<- svm(Species~.,data=iris, kernal="sigmoid",gamma=1,scale=FALSE)
plot(svm.fit,iris,SepalLength~PetalLength)

```

trying different kernels wouldn't affect the performance much until we tune out the best model.

## Answer 3
To improve the performance we can use tune function to determine the best cost and gamma range and then choose bestmodel to classify.

```{r}
tune.out<- tune(svm,Species~.,data=iris,kernal="linear",ranges=list(cost=c(0.001,0.01,0.1,1,5,10,100)))
bestmodel=tune.out$best.model
summary(bestmodel)
```

```{r}
svm.fit<- svm(Species~.,data=iris, kernal="linear",cost=1,scale=FALSE)
plot(svm.fit,iris,SepalLength~PetalLength)
```

```{r}
svmfit1<-svm(Species~.,data=iris,kernal="radial",gamma=0.25,scale=FALSE)
plot(svmfit1,iris,SepalLength~PetalLength)
```

By tuning out the best model we can see the improvement in classification