A Hybrid Approach to Feature Selection and Generation Using an Evolutionary Algorithm

Abstract

Genetic algorithms proved to work well on feature selection problems where the search space produced by the initial feature set already contains the hypothesis to be learned. In cases where this premise is not fulfilled, one needs to find or generate new features to adequately extend the search space. As a solution to this representation problem we introduce a framework that combines feature selection and generation in a wrapper based approach using a modified genetic algorithm for the feature transformation and an inductive learner for the evaluation of the constructed feature set. The basic idea of this concept is to combine the positive search properties of conventional genetic algorithms with an incremental adaptation of the search space. To evaluate this hybrid feature selection and generation approach we compare it to several feature selection wrappers both on artificial and real world data.