How can I restrict the number of features for testing set in SVM classification with linear kernel?

Question

Nikita Mirchandani on 10 Jan 2021

0
Link

Direct link to this question

https://au.mathworks.com/matlabcentral/answers/712963-how-can-i-restrict-the-number-of-features-for-testing-set-in-svm-classification-with-linear-kernel

Answered: Aditya Patil on 17 Feb 2021

I have trained a dataset with linear SVM with 300 features. However, I want to restrict the number of features to 150. How can I find which features are most critical to classification?

0 Comments
Show -2 older commentsHide -2 older comments

Sign in to comment.

Sign in to answer this question.

Answer 1

Aditya Patil on 17 Feb 2021

0
Link

Direct link to this answer

https://au.mathworks.com/matlabcentral/answers/712963-how-can-i-restrict-the-number-of-features-for-testing-set-in-svm-classification-with-linear-kernel#answer_625422

Generally, you can't determine feature importance in SVM, unless a linear kernel is used. Refer following answer for more information. It's recommended to use feature extraction or dimensionality reduction techniques instead of SVM. https://se.mathworks.com/matlabcentral/answers/406577-how-can-i-determine-feature-importance-of-an-svm-classifier