[R] repeated execution of svm(e1071) gives different results, if weights are use and if probability = TRUE is set

[Juergen Rose]

Am Mittwoch, den 23.02.2011, 18:12 +0100 schrieb Juergen Rose:
> class.weights=Wts,

I have just seen, that me last code was not complete. I try it once
more:

> library(e1071)> data(Glass, package = "mlbench")
> index <- 1:nrow(Glass)
> testindex <- sample(index, trunc(length(index)/5))
> testset <- Glass[testindex, ]
> trainset <- Glass[-testindex, ]
> datatrain <- subset(trainset,select=-Type)
> classestrain  <- subset(trainset,select=Type)

If I now repeat:

> model <-
svm(datatrain,classestrain,type="C-classification",kernel="radial",cost=1)
> pred.test=predict(model,datatest);  table(pred.test,t(classestest))


I allways get the same result:

pred.test  1  2  3  5  6  7
        1 12  1  4  0  0  0
        2  0 11  3  2  0  0
        3  0  0  0  0  0  0
        5  0  0  0  1  0  0
        6  0  0  0  0  2  0
        7  0  0  0  0  0  6

Also if I repeat:

> model <-
svm(datatrain,classestrain,type="C-classification",kernel="radial",cost=1,probability=TRUE)
> pred.test=predict(model,datatest,probability=TRUE);
table(pred.test,t(classestest))

I get the same result as above.

But if I set:

> Wts <- 1.0/table(Glass$Type)

and if I then repeat:

> model <-
svm(datatrain,classestrain,type="C-classification",kernel="radial",cost=1,class.weights=Wts,probability=TRUE)
> pred.test=predict(model,datatest,probability=TRUE);
table(pred.test,t(classestest))

each attempt provides a different result:

> model <-
svm(datatrain,classestrain,type="C-classification",kernel="radial",cost=1,class.weights=Wts,probability=TRUE)
> pred.test=predict(model,datatest,probability=TRUE);
table(pred.test,t(classestest))
         
pred.test  1  2  3  5  6  7
        1  0  3  3  3  2  6
        2 12  9  4  0  0  0
        3  0  0  0  0  0  0
        5  0  0  0  0  0  0
        6  0  0  0  0  0  0
        7  0  0  0  0  0  0
> model <-
svm(datatrain,classestrain,type="C-classification",kernel="radial",cost=1,class.weights=Wts,probability=TRUE)
> pred.test=predict(model,datatest,probability=TRUE);
table(pred.test,t(classestest))         
pred.test  1  2  3  5  6  7
        1 12 12  7  3  2  6
        2  0  0  0  0  0  0
        3  0  0  0  0  0  0
        5  0  0  0  0  0  0
        6  0  0  0  0  0  0
        7  0  0  0  0  0  0
> model <-
svm(datatrain,classestrain,type="C-classification",kernel="radial",cost=1,class.weights=Wts,probability=TRUE)
> pred.test=predict(model,datatest,probability=TRUE);
table(pred.test,t(classestest))         
pred.test  1  2  3  5  6  7
        1 12  9  6  0  1  0
        2  0  3  1  3  1  6
        3  0  0  0  0  0  0
        5  0  0  0  0  0  0
        6  0  0  0  0  0  0
        7  0  0  0  0  0  0
> model <-
svm(datatrain,classestrain,type="C-classification",kernel="radial",cost=1,class.weights=Wts,probability=TRUE)
> pred.test=predict(model,datatest,probability=TRUE);
table(pred.test,t(classestest))         
pred.test  1  2  3  5  6  7
        1 12 12  7  3  2  6
        2  0  0  0  0  0  0
        3  0  0  0  0  0  0
        5  0  0  0  0  0  0
        6  0  0  0  0  0  0
        7  0  0  0  0  0  0
> model <-
svm(datatrain,classestrain,type="C-classification",kernel="radial",cost=1,class.weights=Wts,probability=TRUE)
> pred.test=predict(model,datatest,probability=TRUE);
table(pred.test,t(classestest))         
pred.test  1  2  3  5  6  7
        1 10 11  3  3  0  3
        2  2  1  4  0  2  3
        3  0  0  0  0  0  0
        5  0  0  0  0  0  0
        6  0  0  0  0  0  0
        7  0  0  0  0  0  0

What I am doing wrong?