Different metaheuristic strategies to solve the feature selection problem

  • Authors:
  • Silvia Casado Yusta

  • Affiliations:
  • Department of Applied Economics, University of Burgos, Fac C. Económicas y Empresariales, C/Parralillos s/n, Burgos 09001, Spain

  • Venue:
  • Pattern Recognition Letters
  • Year:
  • 2009

Quantified Score

Hi-index 0.10

Visualization

Abstract

This paper investigates feature subset selection for dimensionality reduction in machine learning. We provide a brief overview of the feature subset selection techniques that are commonly used in machine learning. Different metaheuristic strategies are proposed to solve the feature selection problem - GRASP, Tabu Search and Memetic Algorithm. These three strategies are compared with a Genetic Algorithm (which is the metaheuristic strategy most frequently used to solve this problem) and with other typical feature selection methods, such as Sequential Forward Floating Selection (SFFS) and Sequential Backward Floating Selection (SBFS). The results show that, in general, GRASP and Tabu Search obtain significantly better results than the other methods.