FEATURE SELECTION FOR FUZZY CLASSIFIERS USING THE RANKING AND CROSS-VALIDATION
Ilya A. Hodashinsky, Marina B. Bardamova
Tomsk State University of Control Systems and Radioelectronics
The paper compares the effectiveness of modifications of the shuffled frog leaping algorithm that allow metaheuristics to function in the binary search space. Methods based on modified algebraic operations, transformation functions and fusion operations, as well as their combinations, have been tested for the task of selecting features in the fuzzy classifier. The SVC2004 data set containing a large number of features for user authentication based on dynamic handwritten signature characteristics was used in the experiment.
fuzzy classifier, shuffled frog leaping algorithm, metaheuristics, feature selection, binarization