Memetic feature selection: Benchmarking hybridization schemata

Abstract

Feature subset selection is an important preprocessing and guiding step for classification. The combinatorial nature of the problem have made the use of evolutionary and heuristic methods indispensble for the exploration of high dimensional problem search spaces. In this paper, a set of hybridization schemata of genetic algorithm with local search are investigated through a memetic framework. Empirical study compares and discusses the effectiveness of the proposed local search procedure as well as their components. © 2010 Springer-Verlag.