An improved genetic salp swarm algorithm with population partitioning for numerical optimization

The metaheuristic method is effective in solving complex optimization problems. Among these methods, the Salp Swarm algorithm (SSA), inspired by the sailing and foraging behavior of the deep-sea salps population, has proven to be an effective method. However, in practice, SSA is prone to the problems of reduced population diversity and falling into local minima. In order to solve this problem, this paper combines several new techniques and introduces an improved version of SSA called Genetic Salp Swarm Algorithm (GSSA). Specifically, GSSA generates an initial population using a chaotic dyad-based learning method, reduces the dimensionality by a population partitioning technique and performs crossover and mutation operations on the reduced subspace, and finally acts on the optimal solution through three mutation operators to prevent the algorithm from stagnating in the local optimum. This novel GSSA algorithm is tested on 23 benchmark function test sets, CEC2017 and CEC2022. The results show that the GSSA algorithm converges faster and has higher accuracy.