SVM classification - normalization in R

Question

I want to work with SVM classification. How is it possible to normalize (or scale) the features per column in my dataset before i use the SVM model?

train <- read.csv("train.csv")
test <- read.csv("test.csv")

svm.fit=svm(as.factor(type)~ ., data=train, core="libsvm",kernel="linear",cross=10, probability=TRUE)

Answer 1

You can use the scale function in an sapply :

scaleddf <- as.data.frame(sapply(train, function(i) if(is.numeric(i)) scale(i) else i))

If your data contains variables with NaN values or with 0 variance, you could first process and subset the original dataset before using the function above.

# get a vector of variables to drop
dropVars <- sapply(train, function(i) {
              if((is.numeric(i) & !any(is.nan(i)) & sd(i) > 0) | is.factor(i) | is.character(i)) TRUE
              else FALSE
              }
# subset test dropping columns that don't fit the criteria
smallerdf <- test[, dropVars]

Then apply the original sapply function above to smallerdf