# Is Recursive Feature Elimination finding best features subset?

On a set of 9 features I have applied Recursive Feature Elimination (RFE) algorithm using SVM estimator, following approach from (1). When requesting a subset of size 1 to be found, then RFE returned feature X.

However, when I trained SVM over each feature individually, I found another feature Y to have higher accuracy than SVM trained over X.

I thought that RFE finds features with the highest accuracy.

Is my understanding of RFE wrong?

As with most greedy processes, the point of RFE is to reduce the computational cost (fitting a model for each of the $$2^m$$ feature subsets), at the cost of perhaps not finding the actual optimum (but hopefully "close enough").