What are some of the pros/cons of SVM?

Pros:

  • Relatively fast computational time due to kernel trick
  • Ability to learn non-linear decision boundaries
  • Performs well with high-dimensional data
  • No assumptions to verify; however, it is recommended to scale data first

Cons:

  • Performance of algorithm is sensitive to the choice of kernel
  • Does not easily produce probability scores; only class labels
  • No direct way to determine variable importance (might have to use model agnostic permutation approach)