Artificial Bee Colony Algorithm with Hybrid Strategies for Many-Objective Optimization

Artificial Bee Colony (ABC) algorithm is a classical Swarm Intelligence Optimization Algorithm (SIOA), which has been widely used to solve various optimization problems. However, these problems mainly focus on single-objective and ordinary Multi-objective Optimization Problems (MOPs). For Many-objective Optimization Problems (MaOPs), ABC shows some difficulties: (1) the selection pressure based on Pareto dominance degrades severely; and (2) it is not easy to balance convergence and population diversity. In this paper, a new Many-Objective ABC variant with Hybrid Strategies (namely HSMaOABC) is proposed to deal with MaOPs. Firstly, the fitness function is redefined based on objective values and cosine similarity to handle multiple objectives. Then, a new selection method is designed on the basis of the new fitness function. In order to enhance convergence, an elite set guided search strategy is utilized for the employed bee stage, and dimensional learning is incorporated for the onlooker bee stage. Finally, a modified environmental selection strategy is employed based on Penalty-based Boundary Intersection (PBI) distance. To evaluate the performance of HSMaOABC, the DTLZ and MaF benchmarks with 3, 5, 8, and 15 objectives are used. Experimental results demonstrate that HSMaOABC obtains competitive performance when compared with nine other well-known approaches.