What should be value of gamma in SVM

svm
r

#1

I am currently doing a problem of classification when the data is not linearly separable and then I have changed the kernel from linear to radial but in this I also have to defined the value of gamma but I am not able to how to decided the value of gamma so it classify the data accurately.

set.seed(1)
x<-matrix(rnorm(400),ncol=2)
x[1:100,]=x[1:100,]+2
x[100:150,]=x[100:150,]-2
y<-c(rep(1,150),rep(2,50))
dat1<-data.frame(x,as.factor(y))
plot(x,col=y)
train=sample(200,100)
svfit=svm(y~.,data=dat1,kernel=“radial”,gama=1,cost=1)

In this I have taken the value of gamma =1 but I want to know how to decided it.


#2

@hinduja1234- you can use cross validation method to check the value of gamma which is best suited for the model.I would suggest you to use tune cross validation checking the value of gamma in model.