Research On Sparrow Search Algorithm Based On Multi-Strategy Collaborative Improvement

Swarm intelligence optimization algorithms are a type of intelligent optimization algorithm based on simulating the collective behavior of a population.They draw inspiration from the swarm intelligence behavior of various biological organisms,such as ant colonies,bird flocks,and fish schools,to simulate the optimal solution of real-world problems.Due to their good global exploration ability and robustness,swarm intelligence algorithms can find the optimal solution in high-dimensional or multi-dimensional spaces,and have been widely used in many fields such as engineering optimization,economics,data mining,and bioinformatics.The sparrow search algorithm is a relatively new swarm intelligence optimization algorithm that uses the predatory behavior of sparrow flocks as heuristic rules to solve optimization problems.The sparrow search algorithm has the advantages of simple structure,fewer control parameters,easy-to-build code framework,and strong adaptability.It also has excellent global development capabilities and is suitable for solving the optimal solution of complex optimization problems.However,it also has shortcomings such as insufficient population diversity in the later stage and easy to fall into local optimum,which limits the application scope of the algorithm.To address these shortcomings of the sparrow search algorithm,this paper proposes two improved methods and conducts multiple comparative experiments using benchmark functions.The main research work is as follows:(1)This study presents a novel normal mutation sparrow search algorithm with lens reversal learning.Initially,the discoverer’s position update is perturbed by the sine mapping to mitigate its inclination to move swiftly towards the origin.Subsequently,the follower model is refined utilizing the normal mutation operator based on the sign function,augmenting their local exploitation capability.Finally,the scaling factor of the lens reversal learning is optimized using the hyperbolic tangent function,and the improved strategy is applied to population initialization to increase population diversity.Experimental results indicate that the enhanced sparrow search algorithm surpasses the original algorithm in all dimensions.(2)In order to further eliminate the performance limitations of the sparrow search algorithm,a multi-strategy cooperative improved sparrow search algorithm is proposed.Firstly,the Iterative mapping and opposition-based learning mechanism are introduced to improve the population structure and enhance population diversity.Secondly,a forgetting decline strategy is used to gradually reduce the number of individuals using the chaotic opposition-based learning strategy,thereby reducing the computational cost.Thirdly,to enhance the adequacy of global search in the solution space,the dynamic weighted spiral search strategy is employed for the discoverer.Fourthly,incorporating a reference frame mechanism into the discoverer’s position update model,which utilizes the previous iteration’s discoverer position as a reference for updating its own position,is proposed to enhance its local exploitation performance.Finally,a follower model is proposed based on the reference mechanism of the spiral search,aiming to enhance mutual learning and communication among individuals and reduce the likelihood of followers getting lost.Additionally,adaptive weights and dynamic selection strategies are applied to the discoverer and follower models to balance their global and local search capabilities and increase the algorithm’s versatility.Experimental results show that the sparrow search algorithm based on multi-strategy improvement has significantly improved optimization performance in all dimensions.The practicality of the improved algorithm is demonstrated by applying it to PID neural network decoupling control.