ADMEWORKS ModelBuilder - Feature Selection - Iterative SVMFS MIXED
This command displays dialog for SVM-based feature selection with mixed criteria. This method is an extension of the “Entropy-based RFE” method. This method uses the entropy of the parameters ranking to decide how many features should be removed. It also needs the “mean threshold” parameter to control the size of the eliminated parameters set at one iteration. It uses many criteria during the feature selection process, and an average value of criterions is taken for elimination at the actual step.
This method brings good results for difficult data sets.