An Improved Artificial Bee Colony Algorithm and Its Taguchi Analysis

Abstract

The artificial bee colony (ABC) algorithm is one of well-known evolutionary algorithms, which has been successfully applied to many continuous or combinatorial optimization problems. To increase further its convergence speed and avoid being trapped in local optimum, this paper proposes an improved ABC algorithm (IABC), which aims to enhance diversification of search at each stage of the ABC algorithm. Firstly, a chaotic mapping rule is established by introducing a chaos operator into the initial position generation rules in order to ensure the ergodicity of initial positions. Then, an isometric contraction parallel search rule is devised, based on which a neighborhood search on initial positions is performed to enhance the convergence speed and the local search ability. Next, a parallel selection strategy is developed by using roulette and reverse roulette simultaneously, which allows selecting poor positions to escape from local optimum. Meanwhile, a global updating mechanism based on gravitational potential field is developed, which can guide the rejection and generation of positions to accelerate the convergence of the algorithm. The computational results show that the IABC can improve the convergence speed and solution quality without falling into the local optimum prematurely. Finally, a further analysis on the IABC is conducted using the Taguchi method, which focuses on the factor level setting related to the following key factors: chaotic mapping rules in the initial position generation rules, isometric contraction parallel search rules, parallel selection strategies and the update threshold in the global updating mechanism. The results display that the optimal combination of factor levels has been achieved in the IABC.