A Comparative Study of Sequential Feature Selection Methods for Support Vector Machine

Jonghwa Kim

erschienen 2007 Technical Report, Institute of Computer Sciecne, University of Augsburg, October 2007


In this paper we investigate existing feature selection algorithms combined with support vector machine (SBS). Two ranking-based algorithms, recursive feature elimination (RFE) and incremental regularized risk minimization (IRRM), and greedy sequential backward search (SBS) are tested by using biosignal dataset which contains 35 features per sample and a total of 25 samples labeled by four emotion classes. The performance of the selection algorithms are compared by considering recognition rates obtained by the leave-one-out validation....