Intelligent multi-objective optimization ABC algorithm based on dynamic weights and average cognitive strategy

Abstract

Traditional artificial bee colony (ABC) algorithms have common shortcomings in solving multi-objective optimization problems, such as limited search ability, susceptibility to getting stuck in local optima, low solution accuracy, and premature convergence. This study innovatively proposes two ABC optimization algorithms: the multi-objective optimization ABC algorithm based on dynamic weight control, and the multi-objective optimization dual-file ABC algorithm based on average cognitive strategy. The former introduces dynamic domain search and adaptive weight factor allocation at the mechanism level to accelerate convergence and suppress premature convergence. The latter enhances coverage balance on complex frontiers through average cognitive position guidance and coordinated updating of elite/individual dual-file. Compared with other multi-objective bee colony algorithms, the research algorithm performed the best. In the dual-objective ZDT1, ZDT2, and ZDT3 testing functions, the standard deviations were only 4.669 × 10−5, 6.254 × 10−5, and 1.625 × 10−4, respectively, demonstrating high stability and accuracy. For the DTLZ series testing functions, the research algorithm had good convergence effect and good solution set distribution. In addition, the complex UF function testing revealed that the research algorithm was more efficient in solving multi-objective problems and had a higher number of optimal solutions. The research algorithm has significant advantages in multi-objective optimization problems, not only improving the accuracy and stability of the solution, but also enhancing the distribution of the solution set. This method provides an effective solution for optimizing complex systems. The successful application of these algorithms is expected to exert a more critical role in the field of intelligent optimization.