ABSTRACT:

Feature choice may be a very necessary part for datamining, machinery learning and pattern recognition. Distance plays a very important role in Support Vector Machines (SVM) theory. Relief-F algorithm solves feature redundancy well but does not guarantee the maximum distance. To overcome this drawback, a feature subset selection algorithm is proposed that takes SVM average distance as estimation rule and sequential forward selection as search strategy. Using public knowledge set acquired from UCI, this algorithm is compared with the Relief-F. The results show that the recognition rate is over Relief-F with smaller selected options under computation quantity tolerant conditions