A novel improvement of particle swarm optimization using an improved velocity update function based on local best murmuration particle

Abstract Improvement of particle swarm optimization (PSO) is relevant to solving the inherent local optima and premature convergence problem of the PSO. In this paper, a novel improvement of the particle swarm optimization is provided to curb the problem of the classical PSO. The proposed improvement modifies the updating velocity function of the PSO, and it uses a local best murmuration particle which is found using the k-means clustering technique. In this contribution, each particle moves towards the global best position by not only using the personal best and global best, but particles are modelled to move in murmuration towards the global best using the personal best, global best and a local best particle known as the local best murmuration particle. The improved model was tested against the traditional PSO and two other variants of the PSO and genetic algorithm (GA) using 18 benchmark test functions. The proposed improvement demonstrated superior exploration abilities by achieving the best optimum values in 15 out of 18 functions, particularly in the multimodal functions, where it achieved the best optimum value in all 6 cases. It also achieved the best worst-case values in 12 out of 18 functions, especially in the variable-dimension functions, where other algorithms showed significant escalation, indicating the proposed improvement’s reliability and robustness. In terms of convergence, the proposed improvement exhibited the best convergence rate in all 18 functions. These findings highlight the impressive ability of the proposed improvement to converge swiftly without compromising accuracy.