A Sparrow Search Algorithm Based on Dynamic Selection of Evolutionary Directions

The sparrow search algorithm (SSA) is a swarm intelligence algorithm with strong search ability and fast convergence, but it may converge early and fall into local extremes in evolution. To address these problems, an improved multi-strategy sparrow search algorithm based on multi-strategy (MSSSA) is proposed. Firstly, we propose the idea of evolutionary direction division based on the change of individual position for the safety state of the producers. Secondly, the improved upper confidence bound (UCB) and epsilon-greedy methods are fused to achieve better evolutionary direction selection for discoverers in different confidence interval ranges. Finally, the mechanism of normal distribution is analysed to construct a standard deviation function about the evolutionary algebra to realize the adaptive position update of the discoverer, and then improve the iteration capability. The results of benchmark experiments show that MSSSA has better mean, standard deviation and optimal solutions compared with other commonly used SSA algorithms. The improved SSA algorithm has better solution quality.