An Enhanced Bacterial Foraging Optimization Based on Levy Flight and Improved Roulette Wheel Selection

Abstract

For the purpose of promoting the convergence and global search capability of the original Bacterial Foraging Optimization (BFO), this paper proposes a novel BFO combined with Levy Flight and an improved Roulette Wheel Selection (LIRBFO). Due to the chemotaxis step length of the original BFO is set as a constant value. There is no balance between global search and local search. This restricts the application of BFO to tackle complex optimization problems. First of all, a random determining chemotaxis step size using Levy distribution is utilized to substitute fixed chemotaxis step size, which makes bacteria search for food through frequent short-distance search and occasional long-distance search and then improves the optimization ability and efficiency of the algorithm. Moreover, to make up errors caused by randomness strategies and maintain good population diversity, we present an improved roulette selection strategy based on evolutionary stability principle and apply it to the process of bacterial reproduction. The experiments are performed on eight benchmark functions to compare with the performance of LIRBFO, Bacterial Foraging Optimization with Levy Flight chemotaxis step size (LBFO) and improved BFO based on Improved Roulette Wheel Selection (IRBFO). The results obtained indicate this proposed algorithm greatly improves the convergence performance and search accuracy of original BFO in most cases.