How to decide the kernal type in SVM



I am working on a data classification problem. There are few doubts I am facing.

  • How to decide that whether our data is Linearly or Non-Linearly separable with examples?
  • How to decide a kernel for our data classification problem?

Thanks in Advance.


@sonam_gupta - I would suggest you to first visualize the data or you can calculate the R^2 which can help you to decide that is the relationship is linear or non-linear.

The kernel identification is very important part because the wrong selection of kernel can underfit or overfit the model.
<img src="/uploads/analyticsvidhya/original/2X/4/4c053083d942cd345629267243a0ab1c827a54e1.png" width=“603” height="130

Hope this helps!



Thanks a lot. I’ll implement the same.