モデルによって最適なカーネルが異なる
> kernel.vec <- c('linear','poly','radial','sigmoid') > scores <- data.frame() > for(word in kernel.vec){ + iris.svm <- svm(Species~.,iris,kernel=word,cross=5) + prediction <- predict(iris.svm,iris[,1:4]) + result <- table(prediction,iris$Species) + score <- sum(diag(result))/sum(result) + score.df <- data.frame(Kernel=word,Score=score) + scores <- rbind(scores,score.df) + } > scores Kernel Score 1 linear 0.9666667 2 poly 0.9533333 3 radial 0.9733333 4 sigmoid 0.8866667
