Abstract:

The sequential forward selection based on classification accuracy (CA-SFS) was proposed by associating sequential forward selection (SFS) with generalized sequential forward selection (GSFS). It varied the value of r in GSFS and employs SVM (support vector machine)as the classifier. The classification accuracy  was taken as a criterion to decide the retention or elimination of features. Simulations showed that CA-SFS performed  well both in selecting fewer features and classifying samples.

Key words: feature selection; support vector machine; classification accuracy; simulation