0
$\begingroup$

I have a dataset which I would like to use to train SVM models. The dataset contains binary variables as well as variables that are in the range [0,1] (i.e. they represent proportions calculated by dividing a ratio-type of measurement by another ratio-type of measurement).

Since the values of all variables/features fall in the range [0,1], do I still need to standardize (z-score) them (i.e. any/both of the binary and proportion ones)?

$\endgroup$
1
$\begingroup$

If all you know is these variables are bounded between 0 and 1, then I'd still z-score them, because they could still have quite different means and variances. If you additionally have reason to believe that those means and variances are also very similar, you might be able to forego standardization. But then again, what is there to be gained by not z-scoring?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.