SVM on Python: After feature selection

Question

I am working on a dataset with ~80 features. After running a feature selection procedure, I would like to focus on 8 selected features using svm.SVC on Python.

I would like to know whether the input to the svm.SVC should be the data of these 8 features only after feature selection or not. (Or is there any other procedure that needs to be applied before the SVM step?)

Of course you select features for use in machine learning algorithm (to not overwhelm it with too many data), so only these 8 features should be feeded to SVM. And may be you need to normalize or standadize feature values before SVM step. And do something with NaN values in your data if you have some. — CrazyElf
@CrazyElf Thanks! Yes, the normalization etc will be applied. — TTZ

Grr Grr · Accepted Answer · 2017-10-31T14:58:01

You should always scale your data before running an SVM. I would point you to Speed of SVM Kernels? Linear vs RBF vs Poly as an example of why you need to scale your data before running the SVM. Otherwise you don't necessarily need to do anything else before building the model.

SVM on Python: After feature selection

1 Answers