MCDMSR: multicriteria decision making selection/replacement based on agility strategy for real optimization problems

Selection/replacements are an indispensable part in evolutionary algorithms (EAs), which are generally based on a single evaluation criterion. However, selections in nature are based on multi evaluation criteria, such as multi-aspect survival criteria of wolves (like running and attacking abilities). To realize more real survival of the fittest in EAs, multicriteria decision making selection/replacement (MCDMSR) is proposed. In fact, there is little research about using multicriteria decision making to improve selection/replacement models in EAs. VIKOR (a multicriteria decision making method) in management science is used in MCDMSR, and multi evaluation criteria of the populations are synthetically analyzed as a radar chart form. By using VIKOR, MCDMSR is able to respond to the population-state change in real time. MCDMSR is characterized by this agility, and this agility is derived from the decision making ability of VIKOR. Moreover, principle analysis and discussions are given to explain the feasibility of multicriteria decision making in the applications of selection/replacements. We provided the applications of MCDMSR in simple genetic algorithm, particle swarm optimization, artificial fish swarm algorithm, and shuffled frog leaping algorithm, by comparing with tournament selection, fine-grained tournament selection, all-individual-guider replacement, CD/RW, constrained-visual-region replacement, group constrained-visual-region replacement, part-individual-guider replacement for 36 benchmarks (i.e., 5 unimodal and 15 multimodal problems in CEC 2013 test suite, and 16 P-Peak problems). The effectiveness, efficiency, and diversity results of MCDMSR were acceptable.