install.packages('ggplot2', dep = TRUE)
install.packages('e1071', dep = TRUE)

#Load data
data("iris")
str(iris)

#Plot sample data
library(ggplot2)
qplot(Petal.Length, Petal.Width, data=iris, color=Species)

#Use SVM
library(e1071)
mymodel <- svm(Species~., data=iris)
summary(mymodel)
plot(mymodel, data = iris, Petal.Width~Petal.Length, slice = list(Sepal.Width = 3, Sepal.Length=4))

#SVM Model evaluation
pred <- predict(mymodel, iris)
tab <- table(Predicted = pred, Actual = iris$Species)
tab

#Calculate missclassification rate
mcr <- 1-sum(diag(tab))/sum(tab)
mcr

#Change kernel to linear kernel
mymodel <- svm(Species~., data=iris, kernel = "linear")
summary(mymodel)
plot(mymodel, data = iris, Petal.Width~Petal.Length, slice = list(Sepal.Width = 3, Sepal.Length=4))

#SVM Model evaluation with linear kernel
pred <- predict(mymodel, iris)
tab <- table(Predicted = pred, Actual = iris$Species)
tab

#Calculate missclassification rate with linear kernel
mcr <- 1-sum(diag(tab))/sum(tab)
mcr

#Change kernel to polynomial kernel
mymodel <- svm(Species~., data=iris, kernel = "polynomial")
summary(mymodel)
plot(mymodel, data = iris, Petal.Width~Petal.Length, slice = list(Sepal.Width = 3, Sepal.Length=4))

#SVM Model evaluation with polynomial kernel
pred <- predict(mymodel, iris)
tab <- table(Predicted = pred, Actual = iris$Species)
tab

#Calculate missclassification rate with polynomial kernel
mcr <- 1-sum(diag(tab))/sum(tab)
mcr

#Change kernel to sigmoid kernel
mymodel <- svm(Species~., data=iris, kernel = "sigmoid")
summary(mymodel)
plot(mymodel, data = iris, Petal.Width~Petal.Length, slice = list(Sepal.Width = 3, Sepal.Length=4))

#SVM Model evaluation with sigmoid kernel
pred <- predict(mymodel, iris)
tab <- table(Predicted = pred, Actual = iris$Species)
tab

#Calculate missclassification rate with sigmoid kernel
mcr <- 1-sum(diag(tab))/sum(tab)
mcr

#Tune svm model
set.seed(123)
tmodel <- tune(svm, Species~., data = iris, ranges = list(epsilon = seq(0,1,0.1), cost = 2^(2:9)))
summary(tmodel)

plot(tmodel)

#Use best svm model
mymodel <- tmodel$best.model
summary(mymodel)
plot(mymodel, data = iris, Petal.Width~Petal.Length, slice = list(Sepal.Width = 3, Sepal.Length=4))

#Best SVM Model evaluation
pred <- predict(mymodel, iris)
tab <- table(Predicted = pred, Actual = iris$Species)
tab

#Calculate missclassification rate with best model
mcr <- 1-sum(diag(tab))/sum(tab)
mcr