Evaluating performance advantages of grouping genetic algorithms

  • Authors:

  • Evelyn C. Brown;Robert T. Sumichrast

  • Affiliations:

  • Department of Business Information Technology, Virginia Tech, 1007 Pamplin Hall 0235, Blacksburg, VA 24061, USA;Ourso College of Business Administration, Louisiana State University, Baton Rouge, LA 70803, USA

  • Venue:
  • Engineering Applications of Artificial Intelligence
  • Year:
  • 2005

Quantified Score

Hi-index 0.01

Visualization

Abstract

The genetic algorithm (GA) and a related procedure called the grouping genetic algorithm (GGA) are solution methodologies used to search for optimal solutions in constrained optimization problems. While the GA has been successfully applied to a range of problem types, the GGA was created specifically for problems involving the formation of groups. Falkenauer (JORBEL-Belg. J. Oper. Res. Stat. Comput. Sci. 33 (1992) 79), the originator of the GGA, and subsequent researchers have proposed reasons for expecting the GGA to perform more efficiently than the GA on grouping problems. Yet, there has been no research published to date which tests claims of GGA superiority. This paper describes empirical tests of the performance of GA and GGA in three domains which have substantial, practical importance, and which have been the subject of considerable academic research. Our purpose is not to determine which of these two approaches is better across an entire problem domain, but rather to begin to document practical differences between a standard off-the-shelf GA and a tailored GGA. Based on the level of solution quality desired, it may be the case that the additional time and resources required to design a tailored GGA may not be justified if the improvement in solution quality is only minor or non-existent.