Multi-Strategy Improvement And Application Research Based On Sparrow Search Algorithm

In order to cope with various complex optimization problems brought about by the rapid development of the technological era,metaheuristic algorithms have emerged.In contrast to traditional mathematical optimization methods,which have limited accuracy,poor generalizability,and long time-consuming,swarm intelligence optimization algorithms,which have the advantages of simple principles,few parameters,and superiority search speed,have attracted widespread attention.Inspired by the behavior of biological communities in nature,swarm intelligence optimization algorithms have been proposed to draw closer to human life as soon as they are proposed.In 2020,two Chinese scholars proposed Sparrow Search Algorithm(SSA),which imitated the behavior of sparrows.With the advantages of good optimal search performance and powerful self-adaptive capability,it was scrambled by researchers in various fields as soon as it appeared.Nowadays,the algorithm has not only made more effective improvement attempts at the theoretical level,but also been successfully applied to practical problems in various fields,opening up new horizons for data prediction,optimization of model training,scheduling of production processes,addressing and path design,and so on.However,as the in-depth research,the shortcomings of the algorithm itself are revealed,including its unstable initialization mechanism,insufficient optimization capability,more adaptive to the case where the optimal solution is at the origin,and incomplete boundary processing mechanism.In this context,this paper makes an exploratory study to address the above shortcomings,aiming to make its own contribution to improving the performance of the algorithm and its generalizability and practicality.The specific work of this paper can be summarized as follows.(1)In order to improve the inadequate distribution of the initialized population and emphasize the contribution of the initial state of the population to the overall optimization process,this paper combines the idea of chaos to apply three different initialization schemes in the improved algorithms.(2)In order to improve the possibility of the algorithm to escape from the region over-controlled by local extremes,the idea of ranking paired is combined with competitive learning and elastic collision principle for perturbing individuals caught in potential control.In addition,a Gaussian random walking strategy is introduced to contribute to the feasibility of increasing its own escape from the existing region.(3)In order to balance the capability of the algorithm in the exploration and exploitation process,a multi-strategy learning mechanism based on customization idea is proposed to develop a special update mechanism for the characteristics of different locations,i.e.,while retaining individuals occupying the best location,the exchange of location information is boosted to enhance the capability of poorer individuals to find the optimal.In addition,combining the characteristics of the strong local search capability of the whale optimization algorithm,an adaptive spiral predation mechanism is designed to tackle the problem of insufficient local ability of SSA.(4)Multiple sets of experiments were designed.By conducting simulation tests on 22 benchmark test functions and CEC2017 test set respectively,the results reflect that the proposed improved algorithm not only obtains significant improvement in the search performance compared with the basic algorithm,but also has strong competitiveness compared with other variants.This demonstrates the effectiveness of the improvement.Furthermore,by solving four classical engineering optimization problems,the value of the improved algorithm in practical applications is demonstrated.This paper makes an attempt to propose a better improvement strategy for the future,and the work has considerable practical significance and research value.