Feature selection with NSGA and GAAM in EEG signals domain

摘要

The paper presents the comparison of two genetic methods that can be used for feature selection, NSGA (Nondominated Sorting Genetic Algorithm) and GAAM (genetic algorithm with aggressive mutation). While the first method is very popular for optimizing multi-objective functions, the second one is a new method that was introduced just two years ago. The comparison was made with a benchmark file from the second BCI Competition (data set III - motor imaginary). The paper compares both algorithms in terms of the accuracy of the classifiers using features coded in the individuals returned by the algorithms. According to the results reported in this paper, GAAM returned feature sets of the higher classification capacity.