Feature importance with many features

Dear MatLab-Community,

I have a question regarding feature importance.

My input matrix has the the following form: 3000 observations and 240 predictors/features which are classified into 2 groups (0/1).

Now I want to rank the predictors by their contribution to the overall accuracy which is 84% according to the SVM algorithm.

What is the best way to do this and how do I make it statistically sound?

Best regards

Daniel

[Ritish Raghav Maram]

I feel MATLAB had failed to address this concern in many similar queries like this. Scikit learn offers a lot more felxibility in "feature importance" for classification models.

Answer 1

[Shishir Singhal]

Please refer to this link, it nicely explained the techniques to figure out the feature importance in classification model like SVM: https://www.mathworks.com/matlabcentral/answers/406577-how-can-i-determine-feature-importance-of-an-svm-classifier

Answer 2

[yanqi liu]

yes，sir，it is feature select、dimension reduce，may be use mRMR to select，use pca to reduce